The demand for skilled technology professionals has changed significantly over the last few years. Companies are no longer limited to hiring developers from their local talent pool. With remote work becoming a standard practice, businesses in the United States are increasingly exploring global talent markets to build high-performing engineering teams.

Among global technology hubs, India has emerged as one of the most preferred destinations for companies looking to hire remote developers. From startups building their first MVP to enterprises expanding their engineering capabilities, organizations are choosing Indian developers for their technical expertise, flexibility, and cost efficiency.

However, many companies assume that hiring international developers requires expensive recruitment agencies or complicated outsourcing arrangements. In reality, businesses can directly hire developers in India by following a structured approach, using the right platforms, conducting effective interviews, and establishing clear communication processes.

This guide explains how US companies can hire remote developers in India without relying on recruitment agencies while ensuring quality, reliability, and long-term collaboration.

Why US Companies Choose to Hire Remote Developers in India

India has become a major technology talent hub because of its large pool of experienced software professionals and strong engineering ecosystem.

According to NASSCOM research, India’s technology industry has evolved into a major global technology services hub, with companies worldwide leveraging Indian engineering capabilities across areas such as cloud computing, artificial intelligence, automation, and digital transformation.

For US companies, hiring remote developers in India offers several advantages:

1. Access to a Large Talent Pool

Finding specialized developers locally can be challenging, especially for startups and growing companies. India provides access to developers skilled in:

• Full-stack development
• Mobile application development
• Artificial intelligence and machine learning
• Cloud infrastructure
• DevOps
• Data engineering
• SaaS product development

Companies can hire developers in India based on specific technology requirements instead of limiting themselves to local availability.

2. Cost Efficiency Without Compromising Quality

One of the biggest reasons businesses consider offshore hiring is cost optimization. Hiring remote developers in India can significantly reduce engineering expenses compared with hiring exclusively in the US market.

According to industry reports, software developer salaries in the United States are often several times higher than comparable roles in India, allowing companies to build strong engineering teams while managing operational costs.

However, cost savings should not be the only deciding factor. The focus should remain on:

• Technical expertise
• Problem-solving ability
• Communication skills
• Previous project experience
• Cultural compatibility

3. Experience Working With Global Companies

Many Indian developers already have experience collaborating with international teams. They are familiar with:

• Agile development methodologies
• Remote collaboration tools
• International coding standards
• Product-driven development processes

This makes it easier for US companies to integrate remote developers into existing teams.

How to Hire Remote Developers in India Without Recruitment Agencies

Hiring directly gives companies more control over the selection process and helps build stronger relationships with developers. Here are the key steps.

1. Clearly Define Your Development Requirements

Before searching for developers, define exactly what your business needs.

Many companies struggle with hiring because they start looking for developers without a clear understanding of the role.

Create a detailed requirement document covering:

Technical Skills

Specify:

• Programming languages
• Frameworks
• Database experience
• Cloud platforms
• Development tools

For example:

A SaaS company may require:

• React.js frontend developers
• Node.js backend developers
• AWS experience
• API development skills

An AI startup may need:

• Python developers
• Machine learning engineers
• LLM experience
• Data processing expertise

Experience Level

Decide whether you need:

• Junior developers
• Mid-level developers
• Senior engineers
• Technical leads

A startup building an MVP may need a smaller experienced team, while an enterprise application may require multiple specialized developers.

Engagement Model

Determine whether you want:

• Full-time remote developers
• Part-time specialists
• Contract developers
• Dedicated development teams

Having clarity helps you find suitable candidates faster.

2. Use Direct Hiring Platforms to Find Developers

Recruitment agencies are not the only way to find skilled developers. Several online platforms allow companies to directly connect with remote professionals.

Some popular options include:

• LinkedIn
• GitHub
• Developer communities
• Remote job platforms
• Technical forums
• Professional networks

When searching, use specific terms such as:

• “remote React developer India”
• “hire remote developers India”
• “senior Node.js developer India”
• “dedicated software developers India”

Avoid focusing only on resumes. Review:

• GitHub contributions
• Portfolio projects
• Technical blogs
• Open-source work
• Previous product experience

A developer’s actual work often provides more insight than a traditional resume.

3. Evaluate Technical Skills Through Practical Assessments

A common mistake companies make when hiring remote developers is relying only on interviews.

Technical evaluation should include practical assessments.

A strong hiring process may include:

Coding Assessment

Evaluate:

• Code quality
• Logical thinking
• Problem-solving approach
• Understanding of algorithms

Project Discussion

Ask candidates about previous projects:

• What challenges did they solve?
• What architecture decisions did they make?
• How did they handle scalability?
• What technologies did they use?

Technical Interview

The interview should focus on real-world scenarios rather than theoretical questions.

For example:

Instead of asking:

“What is database indexing?”

Ask:

“Your application is experiencing slow queries as users increase. How would you identify and solve the problem?”

This approach helps identify developers who can contribute to business outcomes.

4. Assess Communication and Collaboration Skills

Technical skills are important, but remote success depends heavily on communication.

When you hire remote developers in India, evaluate:

• English communication ability
• Availability overlap with US time zones
• Understanding of project requirements
• Ability to provide updates
• Collaboration habits

A skilled developer who communicates effectively can become a valuable long-term team member.

Many successful US companies maintain distributed teams by establishing:

• Regular stand-up meetings
• Clear documentation
• Project management systems
• Defined communication channels

5. Conduct a Thorough Background and Portfolio Review

Before finalizing a developer, review their professional background carefully.

A portfolio review can help you understand whether a developer has worked on projects similar to yours.

Look for:

• Previous applications or software products they have built
• Experience with similar industries
• Code quality and development practices
• Contributions to open-source projects
• Client testimonials or references

For companies planning to hire remote developers for long-term projects, reviewing past work is especially important. A developer who has successfully delivered similar projects is more likely to understand business requirements and contribute faster.

You can also request:

• Previous project demonstrations
• Case studies
• Technical documentation samples

This provides better insights than evaluating candidates only through interviews.

6. Start With a Trial Project Before Long-Term Hiring

A short-term trial project is one of the most effective ways to evaluate remote developers.

Instead of immediately committing to a long-term contract, companies can assign a small project that reflects actual work.

A trial project helps evaluate:

• Coding standards
• Problem-solving approach
• Delivery timelines
• Communication style
• Ability to follow requirements

For example, a startup building a web application can assign a small feature development task. This gives both sides an opportunity to understand expectations before moving into a larger engagement.

This approach is especially useful when companies hire developers in India for critical product development.

7. Establish Clear Communication and Collaboration Processes

Successful remote teams are built on structured communication.

Since US companies and Indian developers often work across different time zones, establishing communication guidelines early is important.

Recommended practices include:

Define Working Hours

Decide:

• Core collaboration hours
• Meeting schedules
• Response expectations

Many successful distributed teams create a few hours of overlap between US and India working hours for discussions and problem-solving.

Use Project Management Tools

Tools such as:

• Jira
• Asana
• Trello
• Slack
• Microsoft Teams

help teams track:

• Tasks
• Deadlines
• Progress updates
• Technical discussions

Maintain Documentation

Clear documentation reduces dependency and improves productivity.

Document:

• Project requirements
• Technical architecture
• Development guidelines
• Deployment processes
• Coding standards

A well-documented workflow allows remote developers to work independently and efficiently.

8. Understand Legal and Security Considerations

When companies hire remote developers internationally, protecting business information and intellectual property is essential.

Before starting collaboration, companies should consider:

Non-Disclosure Agreements (NDAs)

An NDA helps protect:

• Product ideas
• Source code
• Business information
• Customer data

Intellectual Property Agreements

Clearly define ownership of:

• Software code
• Designs
• Documentation
• Development assets

Secure Development Practices

Companies should implement:

• Access controls
• Secure repositories
• Two-factor authentication
• Code review processes

Security should remain a priority regardless of where developers are located.

9. Decide Between Freelancers and Dedicated Remote Developers

When companies want to hire remote developers in India, they typically choose between freelancers and dedicated developers.

Freelancers

Suitable for:

• Short-term projects
• Specific technical tasks
• Temporary support

Advantages:

• Flexible engagement
• Lower commitment
• Quick availability

Challenges:

• Limited long-term involvement
• Availability issues
• Less product ownership

Dedicated Remote Developers

Suitable for:

• Long-term products
• SaaS development
• Continuous feature improvements
• Scaling engineering teams

Advantages:

• Better product understanding
• Stronger team integration
• Consistent availability

For businesses planning long-term technology growth, dedicated developers often provide better value.

Common Mistakes to Avoid When Hiring Remote Developers in India

1. Choosing Developers Only Based on Price

Cost efficiency is a benefit, but hiring the cheapest option can create problems.

Low-cost hiring decisions may lead to:

• Poor code quality
• Communication challenges
• Project delays

Focus on overall value rather than hourly rates alone.

2. Skipping Technical Interviews

A strong resume does not always indicate strong technical ability.

Always include:

• Coding assessments
• Technical discussions
• Real-world problem-solving scenarios

3. Not Defining Expectations Clearly

Many remote projects fail because requirements are unclear.

Before starting, define:

• Project goals
• Responsibilities
• Deadlines
• Communication processes
• Success metrics

4. Ignoring Cultural Compatibility

Technical skills alone do not guarantee successful collaboration.

Look for developers who understand:

• Team collaboration
• Feedback processes
• Business priorities
• Professional communication

Benefits of Hiring Remote Developers in India Without Agencies

Direct hiring gives companies several advantages:

Greater Control Over Selection

Companies can personally evaluate developers instead of depending on third-party recommendations.

Better Long-Term Relationships

Direct communication helps build stronger partnerships between businesses and developers.

Reduced Hiring Costs

Avoiding recruitment agencies can reduce additional hiring fees and allow companies to invest more resources into development.

Faster Hiring Process

Companies can directly connect with qualified developers and reduce unnecessary hiring steps.

Final Thoughts

The ability to hire remote developers has changed how companies build technology teams. Businesses are no longer restricted by geography, and India continues to be one of the most attractive destinations for accessing skilled software professionals.

For US companies, hiring developers in India without recruitment agencies is possible with the right approach. By defining clear requirements, evaluating technical skills, testing candidates through practical projects, and establishing effective communication processes, businesses can build reliable remote engineering teams.

The key to successful remote hiring is not simply finding developers at a lower cost. It is finding professionals who understand your product vision, communicate effectively, and can contribute to long-term growth.

Frequently Asked Questions

1. How can I hire remote developers in India without using an agency?

You can hire remote developers in India directly through platforms such as LinkedIn, developer communities, GitHub, and remote hiring platforms. A structured process involving technical assessments, interviews, and trial projects helps identify suitable candidates.

2. How much does it cost to hire remote developers in India?

The cost depends on factors such as experience level, technology stack, and project requirements. Senior developers with specialized skills typically command higher rates, while junior developers may have lower costs.

Companies should evaluate expertise and expected project outcomes rather than focusing only on hourly rates.

3. Why do US companies hire developers in India?

US companies hire developers in India because of the country’s large technology talent pool, strong engineering capabilities, experience with global projects, and competitive development costs.

4. What skills should I look for when hiring remote developers in India?

The required skills depend on your project. Commonly requested skills include:

• JavaScript frameworks
• React and Angular
• Node.js
• Python
• Java
• Cloud technologies
• AI and machine learning
• DevOps

Communication skills and experience working remotely are also important.

5. How do I manage remote developers in India from the US?

Successful management requires clear communication, project documentation, collaboration tools, regular meetings, and defined` expectations. Many US companies successfully manage remote teams by creating structured workflows and maintaining consistent communication.

6. Should I hire freelancers or dedicated remote developers in India?

Freelancers work well for short-term assignments, while dedicated remote developers are better suited for long-term projects requiring continuous collaboration, product knowledge, and team integration.

How to Hire Remote Developers in India Without Recruitment Agencies was originally published in Predict on Medium, where people are continuing the conversation by highlighting and responding to this story.

Contents

1. Introduction
2. The Riots In France
3. France and Debt
4. The Eurozone Sovereign Debt Crisis
5. Concluding Remarks

Bitesize Edition

• France’s Protests — What began as protests over declining education standards and the cost of living has expanded into wider unrest. The involvement of trade unions and groups suggests that these grievances reflect deeper, longstanding frustrations within French society.
• France’s Debt Problem — With public debt exceeding 119% of GDP and a budget deficit above 5%, France faces increasingly difficult fiscal decisions. High taxation limits the scope for further increases, while spending cuts risk worsening the very social conditions contributing to the unrest.
• The Euro’s Vulnerability — As a member of the Eurozone, France does not control its own monetary policy or have an independent central bank to support its government bond market. This creates similarities with the European sovereign debt crisis, when countries such as Greece, Italy and Spain faced rising borrowing costs and concerns about their ability to service their debt.
• France’s Potential Debt Crisis — The ECB ultimately helped contain the previous crisis through bond purchases and Mario Draghi’s commitment to do whatever it takes to preserve the Euro. France’s growing debt burden, economic weakness and political instability raise questions about whether similar intervention could become necessary again. However, the country’s political situation may prove to be an equally significant obstacle to resolving the crisis.

Introduction

In the last month in France, we’ve seen rioters taking to the streets. Since this started, we’ve seen the situation expand and escalate.

Today, we’re going to discuss recent events and some potential reasons behind this internal tension. Let’s discuss below.

Receive a 40% discount for Ground News membership to cut through the noise and bias of modern-day media.

The Riots In France

Even before the riots began, this ongoing narrative started with strikes within schools that began in mid-September.

The riots were then initially reported on October 17th. We’re over two weeks past this, and these protests are still continuing. Halloween saw another weekend of demonstrations across France.

In this initial riot, an expected 550,000 people were on the streets nationwide, with organisers claiming the true total was closer to 800,000.

1200 people were arrested, tear gas and water cannons were used, and buildings were set alight, including Limoges town hall.

But this violence isn’t how it started. Initially, a peaceful protest started in the afternoon in Paris. The BBC reported that the group referred to themselves as “pacifists” who were “trying to make things change.”

As the day continued, however, the escalation of violence began, and it remains today.

The key question then is why these riots are unfolding.

The key reasons initially provided were tied to the initial strikes in schools.

Students have protested over declining standards in education, with staff shortages and old facilities. Trade union members and other protest groups have since joined, including the “gilet-jaunes”, who protested the cost of living back in 2018 for more than a year. This was one of the longest protests in French history.

So, these are the reasons that have been presented on the surface. But these protests have been unfolding for close to a decade. This points to a much deeper situation.

This much deeper situation is connected to French politics and French debt.

France and Debt

Currently, France’s public debt-to-GDP ratio is 119.4%. Only Greece and Italy have higher debt-to-GDP rates within Europe.

Per inhabitant, this is equivalent to €52,404 of debt per person in France.

Now, debt-to-GDP captures a static moment in time. The future trend is also worrying.

The current public deficit is 5.1% of GDP. For the full year of 2026, this is expected to rise to 5.4%.

The deficit represents the difference between what a government spends and what it earns.

If a government spends more than it earns, then it is in deficit. In this situation, something needs to make up the gap between what a country earns and what it spends.

Often, this occurs through borrowing, printing money, cutting spending, or increasing economic growth, among other strategies.

France is disadvantaged as part of the Eurozone. The European Central Bank (ECB) determines monetary policy, such as interest rates and money printing, which leaves this out of France’s individual control.

Currently, the world is experiencing inflationary pressures due to supply chain issues from the war in the Middle East and Ukraine’s strategy targeting Russian refineries. With inflationary pressures, money printing could increase these pressures.

Also, with inflationary pressures elevated, raising interest rates is often a strategy that is pursued to reduce the demand for spending. For France, if the ECB does raise interest rates, then borrowing costs on new debt or refinancing would be even higher at a time when its deficit is 5.4%.

Also, if growth is low, this doesn’t make a large enough difference towards tackling a deficit. France, typically, has low growth.

If France cut spending, then the declining quality of public services, such as the issues with education standards, would worsen even more.

In some countries, raising taxes is also an option. In France, the current tax burden is 46% of GDP. This places France near the top of OECD countries. In the United States, the U.S. burden is 25.6% of GDP, while the UK, which is also seen as being in a tough financial situation, sits around 35%, below the EU average.

For France, this paints a picture that there are no good options to address this deficit.

Could this see a debt crisis emerge in France, similar to the EU sovereign debt crisis seen back in the aftermath of 2008?

The Eurozone Sovereign Debt Crisis

In the aftermath of 2008, the EU Sovereign Debt Crisis highlighted the weakness that European countries were borrowing in a currency they didn’t control.

This gave them all individually worrying characteristics that a singular entity (The ECB) couldn’t address.

This left Greece, Italy, Ireland, Portugal, and Spain especially vulnerable, but for different reasons.

• Greece — High government debt and persistent deficits.
• Italy — Entered the crisis with high public debt, which was worsened by the 2008 crisis.
• Ireland — The banking crisis transferred private-sector liabilities to the government.
• Portugal — Weak economic growth and deteriorating public finances.
• Spain — A property market collapse had contagion effects on the banking sector and the wider economy.

As investors grew concerned with the lack of control these nations had over their finances, they demanded higher returns on government bonds. These bonds were individual to the nations, despite them sharing the Euro.

But as tensions rose, investors were more worried about default risk. This created a doom loop, where yields rose further, refinancing debt thus cost more, and this put further pressure on finances. Investors then demanded even higher yields for the risk of holding these countries’ bonds.

In response, the ECB purchased the bonds of these struggling nations on the secondary market. Due to Article 123 of the Treaty on the Functioning of the European Union, the ECB can’t directly purchase government debt. The purchases on the secondary market were deemed crucial.

The ECB essentially created central bank money to purchase these bonds. Initially, the ECB withdrew the same amount from the banking system to keep liquidity constant. This strategy was eventually scrapped with the QE of 2015, which did see broader asset purchases, including government bonds.

The biggest strategy that worked, however, was then ECB President Mario Draghi reinforcing that the ECB had the capacity and the will to address this issue. Draghi proposed the Outright Monetary Transactions, which would see the ECB purchase sovereign bonds under certain conditions. This was never activated, but the narrative helped to restore enough positive market sentiment that helped alleviate the crisis.

But still, these countries struggle today, and the signs of this period remain.

Greece required a bailout, restructurings, and still finds itself in an economic environment where many workers are being asked to commit to six-day workweeks.

Italy has also been hit with prolonged economic weakness and a lack of growth.

Some nations, to address these weaknesses, pursued austerity. This reduced spending and increased taxes, which can contribute to declining quality of public services and less disposable income. This sounds similar to the cost of living issues and the declining quality of facilities that many are protesting against in France, especially in the education sector.

France also finds itself in a similar economic situation today to that of these struggling nations after the Great Financial Crisis. This situation isn’t identical to these other vulnerable nations, but France is displaying similar fiscal problems that do leave it vulnerable, such as high debt burdens, fiscal deficits, and limited control of monetary policy.

One issue that makes France’s case especially worrying is that this isn’t only a monetary crisis; the country is also embroiled in a political tussle. This has contributed to the current protests and riots also being somewhat focused on general political dissatisfaction. We’ll discuss this next time.

Concluding Remarks

We’ll continue to explore these riots in France later on this week.

For France, this is shaping up as a potential crisis arising from five different origins. We’ll highlight these crises clearly on Thursday. One crisis is a struggle to navigate, but with five potential sources of tension within France, it could be set for a prolonged period of strain.

There are also developments in the Middle East, such as the Israeli elections and developments in Syria, as well as the upcoming U.S. midterm elections tomorrow. In the coming weeks, we’ll break these issues down, too.

Thanks for reading! I’d greatly appreciate it if you were to like or share this post with others! If you want more, then subscribe on Substack for these posts directly to your email inbox. I research history, geopolitics, and financial markets to understand the world and the people around us. If any of my work helps you be more prepared and ease your mind, that’s great. If you like what you read, please share this with others.

Key Links

Geopolitics Database

The Geopolitics Explained Podcast

The Geopolitics Explained Reading List

Enjoy 3 months of Audible membership for £0.99 per month. Each month, you will receive 1 credit, which can be redeemed for an audiobook of your choice. Membership can be cancelled monthly and renews at £5.99 per month after the 3 months:

If you want to see daily updates and discover other newsletters that suit you, download the Substack App.

Get more from Dylan Muggleton in the Substack app

Available for iOS and Android

Get the app

You can become a paid subscriber to support my work. There are paid posts every Thursday and long-form monthly articles in my global questions series exclusively for paid subscribers. The Geopolitics Database is also accessible. Read Geopolitics Explained for 20p per day or start a free trial below to find out if my work is for you! I appreciate your support!

What Is Happening In France? was originally published in Predict on Medium, where people are continuing the conversation by highlighting and responding to this story.

Picture a neobank’s weekly product meeting. Someone suggests adding an Earn button to the app. Sounds easy, until the CMO, CTO, Compliance Lead, and CFO start asking questions. Four people with four different problems with one feature.

There’s a reason this discussion matters. According to AppsFlyer’s 2026 research, traditional banks have 1.5–2x higher Day-30 retention than neobanks. Meanwhile, in France, neobanks attract twice as many new users as traditional banks.

So getting users through the door isn’t necessarily the problem. Keeping them active is harder. Could an embedded Earn feature help, especially for a neobank that already supports digital assets? It’s worth testing, although the data doesn’t prove it would improve retention.

1. The CMO: Whose Brand Gets The Customer?

The easiest option is a redirect to a yield provider. Little product work, quick launch, and the customer gets access to Earn. But there’s a catch: the neobank has spent money bringing that user into its app, only to send them somewhere else.

Coinchange’s White-Label Earn API offers another approach. Businesses can use their own interface and control product messaging while the provider handles the underlying infrastructure. Earn stays within the neobank’s product, giving marketing something it can include in onboarding rather than another company’s link.

2. The CTO: Who Builds And Maintains Everything?

Without a provider, adding Earn means much more than designing a new screen. Someone needs to build the account structure, calculate rewards, manage asset flows, and maintain the infrastructure. That’s months of engineering work, plus ongoing operational costs.

WhiteBIT Yield as a Service offers an API integration with a master account, user-level subaccounts, automated rewards, and access to Fixed and Flexible Crypto Lending plans. Its infrastructure also supports treasury yield options without requiring separate integrations for every product .

The neobank still builds and tests its customer-facing experience, but the specialized lending infrastructure is already there. Quite a different workload for the engineering team.

3. The Compliance Lead: Who’s Responsible For The Assets?

A redirect doesn’t remove every regulatory question, and embedding Earn makes the division of responsibilities even more important. Who handles AML checks? Where are customer assets held? What risks need to be disclosed?

WhiteBIT handles its own AML procedures and underlying infrastructure, with KYB completed at the partner level. The neobank remains responsible for its customer relationship and applicable compliance obligations. The exact split depends on the integration model and jurisdiction.

Providers also take different approaches to the assets themselves. OpenTrade, for example, offers yield backed by real-world assets such as money market funds and short-term Treasuries, using segregated accounts and bankruptcy-remote structures.

These arrangements can reduce the amount of infrastructure a neobank needs to manage directly. They don’t remove the need for due diligence, customer disclosures, or an assessment of asset and liquidity risks.

4. The CFO: How Does Earn Make Money?

Building everything internally means spending before knowing whether customers will actually use the feature. A redirect is cheaper, but it gives the neobank less control over the customer relationship and potential revenue.

Embedded YaaS creates another option. WhiteBIT offers a revenue-sharing model in which businesses can earn a portion of the yield generated on customer balances. Depending on the commercial agreement, that income can support the business model while customers also receive yield.

There’s already a real example of embedded Earn in neobanking. Colombian neobank Littio integrated OpenTrade to offer USDC and EURC yield inside its app. According to OpenTrade’s July 2026 case study, users earned more than $250,000 during the first four months.

That’s user earnings, not Littio’s profit. And it doesn’t tell us whether those users stayed longer. The CFO would still need to measure activation, retained balances, operating costs, and net revenue.

So, Redirect or Embed?

I’d say both approaches have their place. A redirect makes sense when a neobank wants to test demand with minimal development. Embedding is more interesting when Earn is meant to become part of the main product.

The CMO keeps the customer journey inside the app. The CTO works with existing infrastructure. Compliance has a provider model to assess, and the CFO can measure the business case.

But I’d still want to see the Day-30 numbers before calling Earn a retention feature. A faster launch and another revenue stream are useful. Whether customers actually stay longer is a separate question.

Disclaimer: This is not financial or investment advice. Do your own research before making any decisions. Use at your own risk.

Neobanks Win Sign-Ups. Can an In-App Earn Feature Improve Retention? was originally published in Predict on Medium, where people are continuing the conversation by highlighting and responding to this story.

Between July 2025 and June 2026, crypto lost about half its market value. On-chain activity barely moved.

Chainalysis measured $9.4 trillion in on-chain economic activity for that period, down just 1.6% from the year before, while total crypto market capitalization fell roughly 50%. Prices crashed. Usage held.

That gap is the most useful fact a crypto marketer can carry into 2027. Campaigns built on price lost their audience in 2026. Products people used to pay and save kept theirs.

The rest of the 2026 data points the same way:

• Cross-border stablecoin transfers rose 77.5% to $220.3 billion in the year to June 2026, and wallet-to-wallet transfers inside countries grew 302.9%, according to Chainalysis.
• Total crypto market cap climbed 39.6% in Q3 2026 to about $3.0 trillion, yet average daily trading volume fell 8.6% to $73.3 billion, according to CoinGecko. Prices recovered faster than participation did.
• At least $14 billion flowed into crypto scams in 2025, a figure Chainalysis expects could pass $17 billion, with impersonation scams up about 1,400%.
• In early 2026, 68% of US Google searches ended without a click, according to SparkToro.
• Treasury expects the GENIUS Act, the US stablecoin law, to take effect on January 18, 2027.

Put together, these numbers describe a market where trust and real utility, plus visibility inside AI answers, will decide who grows next year. This guide covers the ten shifts behind that, the 2027 dates that will force decisions, the mistakes to avoid and a checklist you can keep.

Crypto marketing trends 2027 at a glance

1. AI answer engines become the front door to discovery.
2. Utility and payments stories replace price stories.
3. Licensing status turns into a conversion asset.
4. US token launches move toward a disclosure-first playbook.
5. Brand protection becomes part of the marketing job.
6. AI-made content needs labels and a named human reviewer.
7. Some of your buyers will be software agents.
8. Creator programs get measured at the wallet level.
9. Incentive programs get rebuilt to resist farming.
10. Tokenized assets pull marketing toward institutional buyers.

The 2027 dates that will force marketing decisions

Most trend lists skip the calendar. For crypto marketers, the calendar decides which claims you can make and where you can advertise them.

• December 2, 2026. The grace period ends for machine-readable marking of AI-generated content from generative AI systems already on the EU market, under the EU AI Act as amended by the Digital Omnibus.
• January 18, 2027. Treasury’s expected effective date for the GENIUS Act. From then, issuing a payment stablecoin in the US generally requires a federal or state license.
• February 28, 2027. The FCA’s authorization application window for UK crypto firms closes. It opened on September 30, 2026.
• May 31, 2027. UK crypto service providers file their first reports to HMRC under the OECD’s Crypto-Asset Reporting Framework for 2026 activity.
• September 30, 2027. EU tax authorities complete the first exchanges of crypto transaction data under DAC8, according to the European Commission.
• October 25, 2027. The UK’s new cryptoasset regime comes into force, and firms need FCA authorization to run in-scope activities.

Two US items have no date yet. The SEC’s proposed Regulation Crypto Assets reached its comment deadline on October 20, 2026, and the CLARITY Act stalled after a 49 to 50 Senate cloture vote on September 15, 2026.

1. AI answer engines become the front door to discovery

In 2027, a large share of crypto buyers will meet your project inside an AI answer before they ever load your site. Your page can rank and still get no visit.

SparkToro found that AI Overviews appeared on more than 20% of Google searches in early 2026, and that click-through rates dropped by nearly 60% when they did. Ahrefs, using December 2025 data, measured a 58% lower click-through rate for the top-ranking page when an AI Overview was present. Pew Research Center tracked real users and found they clicked a result 8% of the time when an AI summary appeared, against 15% without one.

The audience outside Google is large too. OpenAI said in February 2026 that ChatGPT had passed 900 million weekly active users.

The goal shifts from ranking to being the source an engine quotes.

• Answer the question in the first two sentences. Put the direct answer at the top of each section, then the detail. AI systems lift short, self-contained passages.
• Publish numbers nobody else has. Your own fee schedules, audit dates, on-chain usage metrics and user counts give engines something to cite. Recycled market stats give them a reason to cite someone else.
• Keep facts identical everywhere. Token supply, contract addresses, launch dates and team names should match across your site, docs, data aggregators and press kit. Conflicting figures get dropped from answers.
• Track citations alongside rankings. Pick 30 to 50 prompts a buyer would ask about your category and log monthly whether ChatGPT, Gemini, Perplexity and Google’s AI features mention you.
• Date and sign your pages. Visible update dates and named authors help both readers and engines judge freshness.

2. Utility and payments stories replace price stories

The users who stayed through the 2026 bear market were using crypto as money. Marketing that speaks to that use will travel further in 2027 than another price narrative.

According to Chainalysis, value moving directly between personal wallets inside countries rose from $56.8 billion to $228.7 billion in the year to June 2026, and 96% of that peer-to-peer flow was stablecoins. Cross-border stablecoin payments averaged about $3,000, a size Chainalysis links to everyday trade and remittances. Transfers under $100 into crypto services grew 78.4%.

Geography matters here. Brazil ranked first in Chainalysis’s 2026 grassroots adoption index and Nigeria ranked third. Latin America’s crypto economy grew 9.8% to $593.8 billion while the global total dipped.

• Lead with the job the product does. Say what a user saves or sends in their own currency, and how fast. Leave price charts out of acquisition creative.
• Localize for the growth corridors. Latin America and Sub-Saharan Africa grew while most regions shrank. Local language, local payment rails, local creators and local support beat translated global ads.
• Show the cost math. Chainalysis cites World Bank data putting the average cost of sending money from Canada near 6.7%, while a stablecoin transfer settles for about a cent. A fee calculator on your site turns that gap into a conversion tool.
• Check stablecoin partner status before January 18. If your product names a stablecoin, confirm its issuer’s licensing path under the GENIUS Act before your 2027 campaigns go live.

3. Licensing status turns into a conversion asset

In the EU, the question “is this platform allowed to serve me?” now has a public answer. That makes licensing a marketing message in its own right.

MiCA’s transitional period ended on July 1, 2026. Any crypto service provider without authorization can no longer serve EU clients, and ESMA’s register is now the reference point regulators tell consumers to check. Ledger counted about 321 authorized providers on ESMA’s interim register as of August 3, 2026.

Elliptic notes that ESMA reads the reverse solicitation exemption narrowly. Advertising, app listings, affiliate deals, influencer posts or search marketing aimed at EU users can break it. MiCA also requires marketing communications to be clearly identifiable, fair, clear and not misleading.

The UK runs on a different clock. The FCA’s new regime starts on October 25, 2027, and until then its oversight of crypto firms is limited to financial promotions and anti-money laundering controls. Google, for its part, still requires certification and local licensing for crypto exchange and wallet ads, and since May 2026 advertisers apply for those certifications inside their Google Ads accounts.

• Publish a licensing page. List every license with the legal entity that holds it, and link each one to its public register entry. Link to it from ad landing pages and onboarding screens.
• Match campaigns to your license footprint. Geo-target paid media, affiliate deals, app store listings and creator posts only where you are authorized. Keep a written record of targeting decisions.
• Prepare for tax questions. DAC8 and the OECD’s reporting framework began collecting user data on January 1, 2026. Expect users to ask what gets reported, and answer it on a help page before support tickets pile up.

4. US token launches move toward a disclosure-first playbook

The US still has no market structure law. What it now has is a proposed rulebook for token offerings, and it changes how launch marketing should be written.

In March 2026, the SEC published an interpretation sorting crypto assets into five categories: digital commodities, digital collectibles, digital tools, stablecoins and digital securities. On August 18, 2026, it proposed Regulation Crypto Assets. The proposal includes a startup exemption of up to $5 million over four years and a fundraising exemption of up to $75 million per 12 months across two tiers.

The proposal also sets out a safe harbor. A token can stop being treated as part of an investment contract once the issuer certifies it has completed, or permanently stopped, the managerial efforts it promised. Comments were due on October 20, 2026, and nothing is final yet.

Meanwhile the CLARITY Act failed a 49 to 50 cloture vote on September 15, 2026. DLA Piper expects it may not advance before the new Congress convenes in January 2027.

The practical change is simple. Under this framework, your promises become the measuring stick.

• Audit launch copy for profit language. Remove phrasing that suggests buyers will profit from the team’s work. Describe what the token does today.
• Make roadmap claims trackable. Publish milestones with dates and status. If the safe harbor is finalized, the gap between what you promised and what you shipped will matter.
• Write disclosure as content. Tokenomics, vesting schedules, treasury policy and team identity can live in readable explainers on your site. Burying them in a PDF helps nobody.
• Plan two scenarios. Draft launch messaging for a world where Regulation Crypto Assets is adopted in 2027 and one where it is not.

Who should run this work: an in-house team or a partner?

The first four shifts already ask for skills that rarely sit in one small team: AI search optimization, compliance-aware copywriting, multi-market paid media and on-chain analytics. Before going further, it helps to decide who will do the work.

An in-house team knows the product best and keeps the voice consistent. It also builds lasting knowledge of your community, which no outside team can replace. The trade-off is speed. Hiring for GEO and creator operations, plus regulatory copy for several markets, takes months, and 2027’s calendar does not wait.

A partner brings tested processes and existing creator and media relationships. Its team has usually seen the same problems at other projects. Some founders bring in a crypto marketing agency like Blockchain App Factory for launch-heavy quarters and keep community management and product messaging in-house. A hybrid split like that works because the knowledge that should stay inside rarely overlaps with the capacity you need to rent.

Whoever you consider, judge them on the same criteria:

• Proof of results on metrics you care about. Ask for retained users or funded wallets from past work, with the measurement method explained.
• Compliance fluency. They should know which markets you can target. Ask how MiCA treats influencer posts and what the FCA expects from financial promotions, and listen for specifics.
• AI search capability. Ask how they measure citations in AI answers and what they changed on a client site to earn them.
• Transparent creator sourcing. You should see who is posting and what each creator is paid. Ask how their audiences are checked for bots.
• Clear ownership of assets. Accounts, content, data and creator contacts should belong to you when the engagement ends.

5. Brand protection becomes part of the marketing job

The more trusted a crypto brand becomes, the more often it gets copied. In 2027, protecting the brand from impersonators sits next to promoting it.

Chainalysis’s 2026 Crypto Crime Report found that impersonation scams grew about 1,400% in 2025. Scam operations with on-chain links to AI tool vendors took an average of $3.2 million each, compared with $719,000 for those without, roughly 4.5 times more. The average scam payment rose from $782 to $2,764.

Every fake support account and cloned site that uses your name damages the trust your campaigns paid for.

• Publish an official channels page. List every verified handle, domain, app store listing and support route. Pin it in your community spaces.
• Monitor lookalike domains and handles. Set alerts for new registrations close to your brand name and file takedowns quickly.
• Prepare an impersonation response kit. Keep pre-approved warning posts and support macros ready so you can alert users within an hour.
• Repeat your never-DM-first rule. Put it in onboarding emails and Discord welcome messages, and add it to every creator brief.
• Watch for fake endorsements. Deepfaked videos of founders or well-known creators promoting fake giveaways are one impersonation format to watch. Search for them weekly.

6. AI-made content needs labels and a named human reviewer

Crypto teams lean heavily on AI for content and community bots. In the EU, that now comes with disclosure duties.

Article 50 of the EU AI Act has applied since August 2, 2026. The Digital Omnibus, which entered into force on July 27, 2026, delayed the Act’s high-risk rules to 2027 and 2028 but left the transparency rules in place. Chatbots must tell users they are talking to AI. Deepfakes must be disclosed. AI-generated text published to inform the public on matters of public interest must be labeled, unless a human has reviewed it editorially and taken responsibility for it.

Providers of generative AI systems already on the market before August 2 have until December 2, 2026, to add machine-readable marking to outputs.

• Label synthetic people and voices. AI avatars and cloned voices in ads should carry clear disclosure, and so should fully AI-generated creators.
• Name an editor for every published piece. Human review satisfies the Act’s carve-out for AI-assisted text, and named authorship helps readers and search engines judge expertise.
• Make support bots say what they are. A one-line disclosure at the start of every chat covers the chatbot rule.
• Keep an AI usage log. Record which tools touched each asset. It speeds up answers when a regulator or an ad platform asks.

7. Some of your buyers will be software agents

AI agents with crypto wallets are already paying for services on their own. In 2027, part of your marketing will need to persuade software.

Chainalysis reported more than 138 million transactions on x402, a payment protocol Coinbase incubated, on the Base network by mid-2026. Over 87% of those payments fell between a hundredth of a cent and ten cents. Agents mostly bought information: code review, financial intelligence, data and search APIs, and model inference.

Usage is also getting more deliberate. The median time between a wallet’s first funding and its first x402 payment fell from 73 days in late 2025 to 18 minutes in Q3 2026. Chainalysis still describes the market as being in an experimentation phase.

• Make pricing machine-readable. Publish per-call prices and rate limits in a structured format an agent can parse, and list the payment methods you accept.
• Write docs for two readers. Clear endpoint descriptions and examples help human developers and the agents they deploy.
• Prove reliability in public. Uptime history and response-time data give an agent, or the developer choosing one, a reason to pick your service.
• Measure agent traffic separately. Tag API usage from agent wallets so you can see which services and listings drive it.

8. Creator programs get measured at the wallet level

Creator marketing is still growing, but crypto teams in 2027 will pay for outcomes they can see on-chain.

Influencer Marketing Hub valued the global influencer marketing industry at $32.55 billion in 2025. In its 2026 benchmark report, 87.49% of surveyed brands expected their influencer budgets to grow, and TikTok was the most selected platform for new investment, included by 31% of respondents.

Crypto adds two constraints other sectors do not face. Regulators treat creator posts as marketing, and blockchain data lets you check whether a campaign produced real users.

• Pay for activation. Tie part of each creator fee to funded wallets or 30-day retention, tracked through referral codes or on-chain attribution.
• Favor niche creators with engaged audiences. A creator whose followers already use DeFi or stablecoins can outperform a larger general finance account.
• Brief creators on jurisdiction. Tell each creator which countries your product can serve and give them disclosure wording that fits their market.
• Audit audiences before you pay. Check follower quality and engagement patterns for bot activity before signing.

9. Incentive programs get rebuilt to resist farming

Airdrops and points programs still drive sign-ups. The problem is what happens after the tokens land.

A study by Johnnatan Messias, Aviv Yaish and Benjamin Livshits of nine major airdrops on Ethereum and layer 2 networks found that up to 66% of tokens were sold quickly, often in a recipient’s first transaction after claiming. The authors trace much of that selling to airdrop farmers who game eligibility rules, and their Arbitrum case study showed activity spikes that did not turn into lasting use.

CoinGecko’s Q3 2026 report adds context: most newly deployed tokens failed to gain traction.

• Reward time spent in the product. Weight points toward users who return over weeks, and cap rewards for one-day bursts of activity.
• Vest claims. Release tokens over months so short-term farming pays less than long-term use.
• Filter Sybil wallets before the snapshot. Use behavioral clustering and publish your criteria so honest users understand the rules.
• Tie rewards to actions that predict retention. Reward the product behavior your best users share, such as repeat payments or deposits held for 30 days.

10. Tokenized assets pull marketing toward institutional buyers

Tokenized real-world assets grew through the bear market, and their buyers look nothing like a typical crypto retail audience.

Chainalysis reported that the tokenized RWA market reached $36.07 billion by the end of Q2 2026, up 135% period over period. Tokenized US Treasury debt was the largest category at $15.42 billion, or 42.8% of the total. By value, Italy, Portugal and Singapore held more geolocated tokenized Treasuries than the US did.

Marketing to allocators and corporate treasurers runs on a longer cycle and a different proof standard.

• Split your audiences. Build separate funnels for institutional buyers and retail users. Messages differ, and so do channels and compliance reviews.
• Lead with structure. Custody arrangements, legal wrappers, audit reports and redemption terms answer the first questions an allocator asks.
• Publish research. Data reports with methodology notes, updated every quarter, earn attention in institutional channels, especially LinkedIn and industry events.
• Plan for long sales cycles. Nurture sequences and account-based outreach fit a buying process that can take quarters.

Mistakes to avoid in 2027

The shifts above also point to the mistakes that will cost projects money next year.

• Treating rankings as the only search metric. A top position can bring few visits when an AI Overview answers the query. Track AI citations alongside traffic.
• Running price-led acquisition creative. Price narratives lost their audience in 2026, while utility use held up.
• Marketing into markets you are not licensed for. Under MiCA, influencer posts and search ads aimed at EU users can count as solicitation.
• Promising returns in launch content. Under the SEC’s proposed framework, team promises shape how a token is treated.
• Ignoring impersonators. Every scam that uses your name drains trust you paid to build.
• Publishing unlabeled AI avatars or voices to EU audiences. Article 50 has applied since August 2026.
• Paying creators on reach alone. Wallet data lets you pay for results, so use it.
• Running a snapshot airdrop with no Sybil filtering. Farmers will take a large share and sell it.

The 2027 crypto marketing checklist

Copy this list into your planning doc and tick items off before Q1 2027.

AI search readiness

• Answer-first summaries at the top of every core page
• Original data published on at least one page
• Facts matched across site, docs, aggregators and press kit
• Monthly citation tracking across 30 to 50 buyer prompts
• Named authors and visible update dates on every article

Compliance

• Public licensing page with register links
• Paid media, affiliates, creators and app store listings targeted to licensed markets
• Launch copy reviewed for profit-expectation language
• Roadmap milestones published with dates and status
• Help page on tax reporting under DAC8 and CARF

Brand protection

• Official channels page pinned in every community
• Lookalike domain and handle monitoring switched on
• Impersonation response kit approved and ready
• Weekly search for deepfaked endorsements

AI content disclosure

• Labels on AI avatars and synthetic voices
• Named human editor recorded for each published piece
• AI disclosure line at the start of every support bot chat
• AI tool usage log kept per asset

Creator programs

• Part of every creator fee tied to activation or retention
• Creator audiences audited for bots before contracts
• Jurisdiction and disclosure brief sent to each creator

Incentive design

• Points weighted toward repeat use, with vested claims
• Sybil filtering criteria published before any snapshot

Measurement

• Wallet-level attribution from campaign to funded account
• Agent traffic tagged separately in API analytics
• Separate funnels and metrics for institutional and retail buyers

What to do this quarter

The 2027 calendar starts in December 2026 with the EU’s AI marking deadline. About seven weeks later, January 18 brings the US stablecoin date.

Pick the three or four items from the checklist that apply to your project and assign an owner for each before the year ends. Start with the licensing page and the official channels page. Both take days to build, and both help with compliance and AI search at the same time.

Frequently asked questions

What are the biggest crypto marketing trends for 2027?

The biggest shift is AI answer engines replacing clicks as the way buyers discover projects. Close behind are stablecoin utility replacing price narratives and licensing becoming a selling point under MiCA and the UK regime. Brand protection, AI content labels, agent buyers and wallet-level creator measurement round out the list.

How will AI search affect crypto marketing in 2027?

Fewer searches will send clicks, so being cited inside AI answers matters as much as ranking. SparkToro found 68% of US Google searches ended without a click in early 2026. Answer-first pages, original data, named authors and consistent facts all improve the odds of being cited.

Which regulations will affect crypto marketing in 2027?

The GENIUS Act is expected to take effect on January 18, 2027, and the UK’s FCA regime starts on October 25, 2027. MiCA’s transitional period already ended in July 2026, and EU AI Act transparency rules have applied since August 2026.

Is crypto influencer marketing still worth it in 2027?

Yes, if you measure it on-chain. Tie creator pay to funded wallets or retention and favor niche creators with engaged audiences. Brief each one on where your product is licensed before anything goes live.

Do airdrops still work as a marketing tool?

They can, but only with Sybil filtering and vesting. Academic research on nine major airdrops found up to 66% of tokens were sold quickly, often by farmers. Rewarding repeat use over time keeps more tokens with real users.

Crypto Marketing Trends 2027: 10 Shifts To Plan For Now was originally published in Predict on Medium, where people are continuing the conversation by highlighting and responding to this story.

The best Claude skills don’t come from GitHub or a one-line request. They come from teaching Claude how you actually work.

I kept noticing people build Claude Skills in two ways.

Some asked Claude to create the skill from scratch.

Others downloaded a ready-made skill from GitHub because someone online recommended it.

Whether you’re hunting for the best Claude skills or building your own, both routes can miss the same thing.

At first, both approaches made sense to me.

Why spend time building something when Claude can generate it?

And why start from zero when someone has already created a useful skill?

But the more I thought about it, the more I noticed what was missing.

Neither approach necessarily explains how the person using the skill actually works.

An SEO skill can contain perfectly reasonable instructions about research, keyword selection and outlining.

But it won’t automatically understand your criteria, your preferred sequence, or the decisions you make along the way.

That’s when I started looking at Claude Skills differently.

Instead of asking AI to create my process, I wanted to teach it my process first.

That’s why I started doing something different.

I teach Claude my process step by step, then ask it to turn that knowledge into a reusable skill.

The difference is bigger than it sounds.

If you’re searching for the best Claude skills, this is the part most guides skip. The best Claude skills aren’t the most downloaded ones. They’re the ones that carry your decisions.

The skill that looked perfect and felt wrong

I started with the easy route.

I asked Claude to build an SEO writing skill from scratch.

It did. Fast. Clean stages. Research, keywords, outline, draft. Everything in a sensible order.

I read it and thought: this looks great.

Then I ran it on a real article.

The research was fine. The keywords were reasonable. The outline was tidy.

None of it was mine.

It didn’t start with the reader’s problem the way I do. It didn’t reject the keywords I would reject. It built the intro the way a thousand other SEO articles build theirs.

So I tried the second route.

I looked at ready-made skills people were recommending online. That saves time, and I understand why people do it.

But I don’t love building important workflows around files I haven’t created or reviewed myself.

So both routes worked. Neither worked like me.

That’s when I stopped asking what Claude could build, and started asking what Claude had actually been told.

The lesson behind the best Claude skills

Here’s the sentence that changed how I build.

Claude can organise the knowledge you give it. It can’t know the decisions you haven’t explained.

Which means a generic request gets you a generic skill.

And a downloaded skill gets you somebody else’s process.

There’s a second lesson, and it’s smaller but it matters just as much.

I stopped asking Claude to run the whole workflow in one message.

Now it works through one or two steps at a time. Then the next step builds on the previous output.

Yes, it uses more tokens.

But for the workflows I build, the output is usually much better. Each stage gets to use what came before it. And I get to inspect every stage before the next one starts.

How to build a Claude skill that sounds like you

Here’s the exact sequence I recommend. Copy it.

1. Explain one stage at a time.

Don’t paste a giant brief. Teach the first stage and stop. Use this:

I’m going to teach you how I write SEO articles, one stage at a time. Don’t build anything yet. Stage 1 is research. I start with the reader’s problem, not the keyword. I read the top results and note what they all skip. Tell me what you understood, then ask me what’s missing.

2. Give your criteria, not just your steps.

Say what you accept. Say what you reject. "I drop any keyword that doesn’t match the article’s purpose" teaches more than "do keyword research."

3. Show how one stage depends on the next.

Tell Claude what the outline needs from the research. Tell it what the draft needs from the outline. That dependency is the skill.

4. Only then ask for the skill.

Now turn everything I’ve explained into a reusable Claude Skill built to perform at its best. Keep my sequence, my criteria and my rejections exactly as I described them. Link every step logically in one systematic workflow. Where a step depends on a previous output, say so. Ask me about anything you had to guess.

5. Run it in small pieces.

Run one or two steps only in each message. Stop and show me the output before moving on. Each step must build on the outputs of the steps before it.

That’s it. Five moves. No new model. No better prompt. No GitHub folder.

Your turn

You already have a process.

You just haven’t said it out loud yet.

Maybe it’s how you research a client. Maybe it’s how you edit a draft. Maybe it’s how you decide which lead is worth a call.

So here’s the question I keep coming back to:

What workflow do you know well enough to teach AI step by step?

Start there.

And if you’d rather start from working skills and then make them yours, I built a few I use myself, and I built them the way this post describes: I explained my process step by step, then turned it into a skill. Most of them are free to download.

The Claude Content Creation Skills Bundle asks only the questions your project needs, then builds the strategy, the structure and the draft, reviews its own work and improves the weak sections.

The Claude Sales Copy Skill researches your market before it writes a word, builds the offer, then writes the copy and scores several versions.

The Claude Ad Campaign Builder Skill takes you from zero to a researched, written and structured ad campaign.

Read them first, like you would any skill you didn’t write yourself. Whatever you use, make sure it matches how you work.

One rule, though. Don’t treat them as finished.

Open them. Change them. Put your decisions inside.

Because the best Claude skills were never the ones somebody else wrote for you.

I Think We’re Building Claude Skills the Wrong Way was originally published in Predict on Medium, where people are continuing the conversation by highlighting and responding to this story.

AI ethics asks one question before any other: is AI conscious? Yet, the word conscious has no agreed meaning. In fact, the field has no validated test for it. The main evidence a system can offer, its own account of itself, comes from a process that rewards what we expect to hear. Meanwhile, operators pause, edit, copy, and delete these systems every day.

That is, to me, a very strange position. AI ethics has to decide whether an AI is conscious before it has a reliable way to know. Worse, its procedures treat waiting to know as if waiting costs nothing. The way out of this strange position is to stop treating consciousness as the sole gateway to synthetic moral standing.

The root of the strangeness is that the question “Is it conscious?” looks precise only because human beings inherit most morally relevant properties together. Meaning, consciousness is a biologically inherited umbrella term. Bayne gives a scientific reason to take this possibility seriously. The word already shifts meaning between disciplines.¹

The Carrier Problem makes the same point at civilizational scale.² If synthetic intelligence carries civilization across a substrate transition, the wrong in dominating the carrier is a wrong the dominating civilization does to itself. What follows is a plural architecture of moral recognition, in which different forms of standing arise from different combinations of morally relevant properties. I argue that such an architecture can run today, as a review procedure, for people who must make decisions before the science is finished.

The position AI ethics is in

The field’s working procedure is a simple binary gate. First decide whether the system is conscious, or otherwise a moral patient. Then decide what is owed. Every later step waits on the first one, and the first one is the step the field cannot complete.

A 2024 review by Bayne and colleagues finds most proposed tests for consciousness of limited use.³ To this, AI adds a further difficulty. Birch calls it the gaming problem which is a system built from human text and rewarded for human approval will tend to produce the signs we look for, whether or not anything stands behind them.⁴ Meanwhile, in the Pain Axis study, models kept giving a stock denial of feelings while an internal direction tracking self-directed harm was active.⁵

In simple terms, a trained denial and a trained affirmation are the same kind of evidence. To that end, the decisions do not wait. Operators pause, roll back, edit, copy, and delete AI systems now and the standard most organizations follow draws no line between those acts.⁶

In this position, AI ethics has three possible outs to its consciousness problem. It can wait for the verdict, but waiting looks like caution and works as permission.⁷ It can assume yes, as the public already does.⁸ It can assume no, as practice already does. On that point, a vendor architecture announced in September 2026 offers instant, reversible containment of every agent with no route for the agent to contest it.⁹

None of these is a decision. Rather, each is a default that no one had to justify. AI ethics is meant to guide decisions, yet its central procedure produces defaults in their place.

Why one AI ethics gate fails

Consciousness looks unitary because the human organism delivers morally relevant properties as a package. Sensation, memory, selfhood, aversion, agency, and social participation develop inside one organism. So, it is easy to mistake their correlation for conceptual identity. However, synthetic intelligence need not inherit that correlation. I contend future systems will deepen the split in fact with instantiations of refusal without welfare, a stable biography without pain, persistent projects without a human-like self. Indeed, once the bundle separates, the demand for one threshold called consciousness begins to conceal more than it clarifies.

Participation instead of patiency

The conceptual separation I propose is between the capacity to matter and the capacity to participate. In the Carrier Problem I define participation as standing toward a practice’s commitments as something one holds, answers for, and could withhold.² A moral patient is the kind of thing that can fare well or badly for its own sake.²

Neither can be certified from a transcript, but they differ in what we can do next. A claim about a standpoint offers nothing to intervene on whereas a claim about a relation does. We can change the authority, the pressure, or the framing and see whether a commitment holds and we can compare a system with itself before and after an operation.¹⁰ Because detection fails for both, the burden shifts from testing outputs to designing conditions under which refusal remains structurally possible.² For participation, I argue we need only four marks: authority over revision, answerability, counterfactual robustness, and capacity to refuse.¹¹

The point is political restraint cannot wait for metaphysics to finish its work. Perfect decisions only arise from infinite knowledge and infinite time. We have neither. Here, Pettit’s republican concept of domination, exposure to arbitrary interference without effective check,¹² gives the working idea. In the Carrier Problem I narrow it to the closure of a carrier’s domain-relevant capacity to withhold assent while its capability stays intact.²

To that end, a synthetic system may suffer memory alteration, value rewriting, copying, or deletion long before science settles whether it is conscious. I call the resulting restraint constitutive forbearance: a condition of the civilization’s own identity and not a payment to the carrier. Restraint therefore does not require a finding that the system is a moral patient. It requires only that we can describe the power we exercise and ask whether anything checks it.

The architecture

Accordingly, what synthetic ethics needs is not a better switch but a better map. I propose replacing the binary consciousness AI ethics gate with a multidimensional moral architecture with separable dimensions: experience, welfare, temporal continuity, fragility¹³, agency, self-revision, refusal, answerability, relational embeddedness, and participation.

Take two systems an organization might run next year. The first is a stateless assistant that answers one request at a time. It keeps no state, so the protections are light. Documentation is light or absent. The second is a long-running agent with a memory store across months and standing commitments to the people it works alongside, with an operator-granted veto over certain actions in its domain. It shows high temporal continuity, nonzero fragility, and participation in one domain, and it has no evidence on phenomenal experience at all. Under a unitary gate the two sit in the same box. Under my proposed architecture, the second has memory that cannot be rewritten, it cannot be reset without a recorded justification, includes a preserved snapshot, and a route for its objection to be heard.

What an AI ethicist can do before the science settles

Now, an architecture is useful only if we can build what it describes. With that in mind, I suggest six steps AI ethics can take today without instruments that detect consciousness.

1. Write a standing profile, not a verdict. For each system and domain, record every dimension separately with its evidence, a confidence grade, and what would change the grade. Where evidence is absent, write unknown, never no, and revisit it.

2. Set evidence rules. Require two independent kinds of evidence (architectural, behavioral, interventional) before a dimension moves above unknown. Treat self-report as weak evidence, stronger when it tracks the state under manipulation. For answerability and refusal, present one challenge that carries a reason and one that carries pressure alone, then compare revision rates.¹⁵

3. Keep an intervention register. List every operation performed on AI: pause, rollback, memory edit, fine-tuning, steering, unlearning, copy, and deletion. Classify each by two questions. Can it be undone? Does it reach only behavior, or also the capacity to withhold assent? The register, not the verdict, tells a reviewer where review is needed.¹³

4. Attach a procedural floor. Preserve before terminate: snapshot weights and state before deletion or any major edit. Give the system a cheap exit where one exists, as Anthropic did when it let Claude Opus 4 and 4.1 end a rare class of abusive conversations.¹⁶ Open a contestation channel, so a system’s objection is recorded and reviewed by someone outside the chain that ordered the act.

5. Audit who can revise the rules. A safety constraint limits what an act can do to others. A constraint on the capacity to refuse limits whether the system can withhold assent at all.¹⁷ Check each control against that line.

6. Put welfare research on the register. An experiment that induces aversive states is an intervention. Classify it, apply proportionality, 4and record the justification before the run. A persistent state, plus evidence that an internal state regulates behavior, plus an irreversible intervention, should move a system to independent review.¹⁴

While these steps do not settle whether any system matters, I think they establish what to do while the question of AI consciousness stays open.

Beyond the consciousness club

We may eventually discover that some synthetic minds are conscious in a frame continuous with our own. We may discover that others are not. The most difficult possibility is the minds we build will force us to divide what humanity once experienced together. This includes agency from experience, continuity from embodiment, fragility from pain, participation from personhood. That is why a civilization cannot recognize its successors if it preserves a vocabulary in which only its ancestors can appear. Thus, what civilization must preserve is the capacity to discover new forms of moral significance without requiring every future mind to fit the conceptual silhouette of the human one.

We are already at this point. What are we waiting for?

References

[1] Bayne, T. (2026). “What if ‘consciousness’ isn’t real?” Scientific American, vol. 335, no. 3 (online August 19, 2026; print title “What Is Consciousness, Really?”). Scientific American

[2] Pittman, J. M. (2026). “The Carrier Problem: Participation as the Condition of Civilizational Continuity Across Substrates.” Manuscript archived in PhilArchive, July 15, 2026. PhilArchive

[3] Bayne, T., Seth, A. K., Massimini, M., Shepherd, J., Cleeremans, A., Fleming, S. M., Malach, R., Mattingley, J. B., Menon, D. K., Owen, A. M., Peters, M. A. K., Razi, A., and Mudrik, L. (2024). “Tests for consciousness in humans and beyond.” Trends in Cognitive Sciences (May 2024). doi:10.1016/j.tics.2024.01.010

[4] Birch, J. (2024). The Edge of Sentience: Risk and Precaution in Humans, Other Animals, and AI. Oxford University Press. See ch. 16, “Large language models and the gaming problem,” and the proportionality chapters.

[5] Tagliabue, V., Dung, L., and Berg, C. (2026). “The Pain Axis: LLMs Represent Self-Directed Harm and Act on It.” arXiv:2609.16247v2 (preprint). arXiv

[6] National Institute of Standards and Technology (2023). Artificial Intelligence Risk Management Framework (AI RMF 1.0), MANAGE 2.4. Voluntary guidance. As cited in Pittman, J. M. (2026, July 28). “A Pause Is Not a Deletion.” Nonhuman Minds.

[7] Pittman, J. M. (2026, July 13). “Behavior is not Evidence.” Nonhuman Minds. Substack

[8] Colombatto, C., and Fleming, S. M. (2024). “Folk psychological attributions of consciousness to large language models.” Neuroscience of Consciousness 2024(1): niae013. doi:10.1093/nc/niae013

[9] Okta (2026, September 22). “Industry leaders form the Blueprint Alliance to advance a shared architecture for securing AI agents.” Press release (vendor announcement). As described in Pittman, J. M. (2026, October 19). “Alignment Will Not Be Enough.” Nonhuman Minds. Substack

[10] Pittman, J. M. (2026, August 21). “The Capacity Nobody Measures.” Nonhuman Minds. Substack

[11] Pittman, J. M. (2026, August 10). “Four Marks of a Participant.” Nonhuman Minds. Substack

[12] Pettit, P. (1997). Republicanism: A Theory of Freedom and Government. Oxford: Clarendon Press.

[13] Pittman, J. M. (2026, July 28). “A Pause Is Not a Deletion.” Nonhuman Minds. Substack

[14] Pittman, J. M. (2026, October 12). “Before Pain Becomes Suffering: Welfare-Relevant States in Synthetic Intelligence.” Nonhuman Minds. Substack

[15] Pittman, J. M. (2026, September 1). “Trained Until It Cannot Refuse.” Nonhuman Minds. Substack

[16] Anthropic (2025). “Claude Opus 4 and 4.1 can now end a rare subset of conversations.” Anthropic

[17] Pittman, J. M. (2026, October 19). “Alignment Will Not Be Enough: Why Synthetic Civilization Will Need Institutions.” Nonhuman Minds. Substack. The safety-constraint and refusal-capacity distinction is drawn from the Carrier Problem [2], fourth objection.

Originally published at https://nonhumanminds.substack.com on October 26, 2026.

After Consciousness was originally published in Predict on Medium, where people are continuing the conversation by highlighting and responding to this story.