Understanding HMRC’s AI use in R&D claims, and how to avoid automated enquiries

There’s a growing use of artificial intelligence in the workplace, and the public sector is no exception. It’s more important than ever to understand the challenges of HMRC AI use in R&D claims, and combat R&D enquiry risks.

This discussion demonstrates the growing concerns around HMRC R&D tax claim transparency for AI, whether HMRC must disclose their use of automation in compliance decisions, and whether we’ll see fairer treatment of R&D claimants in future.

The new reality: how does a landmark tribunal put HMRC’s AI use in the spotlight?

A First-tier Tribunal has ordered HMRC to confirm whether it used AI when working on an R&D claim. The ruling overturned an earlier ICO position and set a compliance deadline, ordering HMRC to be more transparent about its use of AI in R&D checks.

The HMRC AI tribunal has raised questions about HMRC’s AI risk checks for R&D, highlighted Frascati Manual challenges and raised speculation about how accurately these methods may be used.

The tribunal has set a precedent for HMRC R&D tax claim transparency regarding AI, compelling HMRC to reveal whether automated systems were used in claim cases. This focus on transparency will shape how future R&D claims are assessed under the merged scheme.

What was the Elsbury v the Information Commissioner case?

Thomas Elsbury (the individual who brought the case) became suspicious that HMRC had used AI in his R&D claim, after em-dashes and American spellings were used in correspondence that didn’t align with the facts.

Thomas Elsbury’s 2025 case against The Information Commissioner UKFTT 915 (GRC) resulted in a ruling that HMRC must disclose if and when Artificial Intelligence (AI) was utilised in R&D cases.

This Freedom of Information Act (FOIA) ruling, seeking to improve HMRC AI R&D claims challenges, gave HMRC 35 working days to either disclose the information or serve a refusal notice. 

Previous to this, HMRC (His Majesty’s Revenue and Customs) has neither confirmed nor denied whether they use AI risk checks for R&D, but their latest annual report and accounts do state that generative AI is generally used across departments to enhance compliance.

Why is the tribunal decision a game-changer for businesses?

Understanding HMRC AI R&D claims challenges can help you tailor your claim towards the process. The goal is to compliantly design submissions that reduce AI risk checks and R&D false positives before they start.

What key questions does the ruling raise about transparency?

The latest light shed on HMRC AI R&D claims challenges has left many with questions about their R&D claim:

• Which AI systems, if any, are used by HMRC and at what stage?
• Does human oversight always control outcomes?
• How will taxpayers be told if automation affected their case?
• What safeguards protect confidential data in R&D Tax Credits for artificial intelligence (AI)?

It’s going to be important for HMRC to answer these questions in a way that assures claimants that their data is safe, their claims are assessed fairly without needless R&D enquiry risk,  and that the claims process is easy to use.

HMRC R&D tax claim transparency and AI: how does this affect businesses?

The HMRC R&D tax claim transparency AI debate has questioned whether taxpayers have the right to know when automation has played a part in claim outcomes. Businesses should make sure all documentation and technical narratives are sufficiently detailed to withstand both human and AI scrutiny.

HMRC AI R&D claims challenges – what risks does your claim face from AI analytics?

AI can spot patterns much faster than a compliance agent, but it can also misread nuance in technical uncertainty, and people can be more trusting of AI’s ability than they should be. Understanding the risks helps you lower AI risk checks for R&D while meeting tax credits guidelines.

What is the risk of AI pattern-matching and what does it mean for your claim?

If your claim’s technical narrative resembles templated reports, or costs sit outside sector norms without explanation, HMRC AI R&D claims tools may rank you higher risk.

R&D enquiry risk has risen. HMRC have been making an increasing amount of compliance checks for ‘high-risk’ claims. These checks increased from 10% in 2023 to 17% in 2024. HMRC have more compliance officers now checking claims, but there is some evidence to suggest these officers are utilising AI risk checks for R&D.

Why might AI cause more HMRC R&D enquiries?

Inadequate supporting evidence, inconsistent information and working in an industry that is under increased scrutiny are all factors that can increase your R&D enquiry risk.

What are the dangers of using AI tools to draft your R&D report?

Drafting tools can speed up notes, but they can be inconsistent, fall short of up-to-date technical knowledge and lack context. Without expert rewriting against DSIT’s Meaning of R&D guidelines and HMRC’s compliance materials, you risk template signals that attract AI risk checks for R&D.

The solution: what proactive strategy will future-proof your R&D claim?

A human-led technical narrative, audit-ready records and clear scoping between routine work and genuine advances remain the safest path. This is also how to prepare for AI risk checks for R&D with confidence.

Step 1: how do you build a robust, human-led technical narrative?

Lead with the uncertainty and the advance sought in the first three lines. Name the hypothesis, failed routes and why public knowledge did not suffice for a competent professional. Use DSIT wording consistently to anchor eligibility and minimise HMRC AI misreads.

Step 2: how do you make contemporaneous records audit-proof?

To avoid HMRC AI R&D claims challenges, create evidence while you work of:

• Time-stamped tickets, bench notes, and design decisions
• Version control and test logs tied to specific uncertainties
• Supplier statements that describe uncertainty-driven work
• Clear cost reconciliations to CT600L and the Additional Information Form

Step 3: how do you differentiate routine development from genuine advances?

Prevent HMRC AI R&D claims challenges by mapping your activities to DSIT’s qualifying vs non-qualifying lists. Exclude routine debugging, cosmetic changes and integration, and show why the core breakthrough was not readily deducible at the time.

How does human expertise compare to AI risk checks for R&D?

Research shows that AI does not truly understand the frontier of knowledge. Experienced consultants translate experiments into eligibility, spot weak signals and reduce HMRC AI R&D claims challenges without diluting accuracy.

What are the limitations of AI in understanding scientific uncertainty?

AI models infer from past text, not from the actual state of science or technology. They can miss dead ends, negative results and boundary conditions that prove scientific and technological uncertainty.

What is the value of an experienced consultant’s sixth sense?

An experienced consultant reviewing an R&D claim will be able to take a balanced view when looking at borderline cases, and make a decision which is fair in the context of the scheme. This nuance is often where AI falls down.

What long-term impact could AI auditing have on the UK’s innovation economy?

HMRC’s plans to use AI risk checks for R&D to improve efficiency, and if this implementation goes smoothly, innovators could see a faster, more accurate R&D claims process. Well-prepared claimants should still benefit from relief designed to support growth and the UK economy. 

How could increased scrutiny affect legitimate claimants?

Expect more HMRC AI R&D claims challenges like targeted questions, stricter use of guidance and closer checks on record-keeping. HMRC’s published approach and Guidelines for Compliance will help you stay compliant.

What might HMRC’s future use of AI look like?

HMRC has deployed AI to improve the speed and accuracy of work, which may mean your claim is processed faster. However, there may be criteria that is flagged more harshly for an enquiry, leading to more HMRC AI R&D claims challenges.

Mini case study: how did a precise claim narrative pass HMRC scrutiny?

A nutraceutical innovator wanted to create more effective supplements that improved absorption. Previous options offered poor bioavailability, showing genuine technical uncertainty. 

Alexander Clifford presented qualifying activities and costs, framed the advance and uncertainties up front, and evidenced results. The claim avoided HMRC AI R&D claims challenges, delivering a client benefit of over £9,700.

AI risk checks for R&D – FAQs

The HMRC AI legal case has raised a few questions.

What exactly is a technological uncertainty in an R&D claim?

It is a limitation in existing science or technology that a competent professional cannot resolve with publicly available knowledge, and which your project attempts to overcome to achieve an advance.

Can I use an AI tool like ChatGPT to write my R&D report?

Use it only for rough notes. Ensure a human expert rewrites to DSIT standards, adds references to experiments and removes templated phrasing that can trigger AI risk checks for R&D.

How can I tell if my R&D claim was reviewed by AI?

Currently, there’s no way to be sure if a previous claim was reviewed by AI. Following a recent tribunal ruling, HMRC may be required to confirm whether and when AI was used, improving transparency for future cases.

What happens if my claim is flagged for an HMRC enquiry?

Respond quickly with project-level narratives, contemporaneous records and clear cost bridges tied to your AIF and CT600L. It’s always good to prepare for an enquiry but keeping thorough documentation of your R&D activities.

Which businesses are most at risk from AI-driven R&D checks?

Claims with generic narratives, unusual cost ratios without explanation, or sectors with higher historic error rates receive more checks. Using HMRC’s Guidelines for Compliance reduces your R&D enquiry risk.

How do AIF and CNF rules interact with AI reviews?

Submitting the AIF is mandatory for new claims and many claimants must file a CNF within 6 months if they have not claimed recently. These forms will contain information that will be reviewed as part of your claim.

What does HMRC R&D tax claim transparency AI mean?

HMRC R&D tax claim transparency AI refers to the growing requirement for HMRC to disclose when and how artificial intelligence is used in assessing research and development tax relief claims.

How Alexander Clifford can assist your claim

At Alexander Clifford, we have a good relationship with HMRC and in-depth knowledge of shaping claims that clearly represent qualifying criteria. This gives you the best chance of preventing HMRC AI R&D claims challenges.

When HMRC enquiries do crop up, we compliantly advocate for the payout of all of your eligible tax credits – offering a strong defence against AI risk checks for R&D.

Benefit from a free consultation and find out more about our no-win, no-fee service by speaking to our tax credits experts.