And why “doing R&D” is no longer the same as “successfully claiming it”

In 2026, the biggest risk to an R&D tax claim in the UK isn’t bad science.
It’s bad process.

The rules haven’t suddenly become unreasonable, but they have become unforgiving. Deadlines are sharper. Forms are mandatory. Evidence expectations are higher. And HMRC scrutiny is far more targeted than it was even two or three years ago.

Here are the top 10 pitfalls that continue to sink otherwise legitimate R&D claims and why avoiding them is now a competitive advantage.

For many companies, if you don’t notify HMRC within six months of the period end, you don’t get a second chance.

No appeal to “but the work qualifies”.
No rescue by a better report later.

This is governance, not innovation and it’s astonishing how many claims still fail before the science is even looked at.

The Additional Information Form is not an admin afterthought. It’s effectively HMRC’s first line of triage.

Weak descriptions, inconsistent figures or missing details don’t just slow things down; they can invalidate the claim entirely.

By 2026, most failed claims fail here.

With legacy SME rules, RDEC, the merged scheme and R&D-intensive provisions all co-existing, getting the period start date wrong by a single day can change:

• Who can claim
• Which costs qualify
• How subcontracting is treated.

This isn’t pedantry, it’s determinative.

HMRC isn’t paying for ambition, disruption, or “industry-leading platforms”.

They’re paying for:

• Clearly defined technological uncertainty
• Genuine attempts to resolve it
• Advancement against a known baseline.

Buzzwords don’t survive enquiries. Evidence does.

Hard work does not equal R&D
Complex projects do not equal R&D
Commercial risk does not equal R&D

If a competent professional could have resolved the issue without experimentation, HMRC won’t see it as qualifying, no matter how impressive the end product looks.

Post-April 2024, entitlement follows who decides to do the R&D and who bears the risk, not simply who pays the invoice.

This is now one of the most common enquiry triggers:

• The wrong party claiming
• Misunderstanding delivery vs decision-making
• Assuming contracts don’t matter.

They do.

Overseas EPWs and subcontractors are now heavily restricted, with only narrow exceptions.

In 2026, “our developers are abroad” isn’t a footnote; it’s often a structural eligibility issue that can materially reduce or eliminate relief.

If staff time, software or consumables are allocated on instinct rather than evidence, the claim is fragile.

HMRC doesn’t need perfect records, but they do expect:

• A rational methodology
• Consistency with payroll and accounts
• A defensible audit trail.

Anything else is a risk multiplier.

Grants don’t automatically kill claims.
But they do change how costs are treated and sometimes which scheme applies.

The biggest mistake is binary thinking:

“We got funding, so either everything is excluded… or nothing is.”
Reality sits in the detail.

By the time HMRC asks questions, the people who understood the work have often moved on.

In 2026, late claims don’t just risk deadlines; they risk losing the ability to reconstruct the truth.

And that’s when good R&D becomes an indefensible claim.

AI can help structure reports.
It can summarise notes.
It can even draft narratives.

But R&D tax relief still lives and dies on context, judgment, nuance and challenge, the things AI consistently struggles with.

You still need people at the helm:

• To interrogate eligibility
• To push back on over-claims
• To protect clients from short-term wins that create long-term exposure.

In 2026, the question isn’t:
“Did you do R&D?”

It’s:
“Can you evidence it, under the right rules, in the right way, at the right time?”