How to make an R&D tax advisory firm visible in AI search
Four things moved the needle in our sweep, in this order: publish a real compliance figure, name a sector specialism with a stated reason, keep scheme content current, and check per engine — never a blended average.
Steen Stones · Reviewed 5 Aug 2026
Publish a real compliance number first. Everything else on this page matters less than that one action, because every recommendation quote in our sweep was built on a specific, checkable figure or process — an enquiry rate, a claim-value average, a sign-off step — never a general trust statement.
1. A specific compliance or track-record figure
“Under 5% HMRC enquiry rate.” “Average claim value over £145,000.” “A lawyer-reviewed sign-off layer.” These are the actual phrases our sweep found doing the work of a recommendation. A general claim of expertise did not appear in a single winning quote we pulled.
2. A named sector specialism with a stated reason
The strongest quote in our data didn't just say a firm specialised in biotech — it said why that mattered: “PhD-level technical understanding combined with tax expertise reduces the risk of HMRC challenging the science behind your claim.” Name the sector and state the specific risk the specialism reduces, not just the label.
3. Content that reflects the current scheme
The SME and RDEC schemes merged for accounting periods starting on or after 1 April 2024, and the Additional Information Form has been mandatory since August 2023. Any page still framed around the old SME/RDEC split, or silent on the AIF, reads as stale to a careful reader and risks an AI drafting tool regenerating outdated guidance from older training data.
4. Check every engine separately, never a blended average
In our sweep, every one of 87 firms shows zero recommendations from Gemini — the same pattern we found independently in a completely different market. ChatGPT and Claude both converted real, different shares of their own mentions for the same firms. Track recommendation rate per engine, or a genuinely working ChatGPT strategy will be invisible inside an average Gemini's flat zero is dragging down.
“There's a difference between showing up and being recommended. AI cannot recommend what it doesn't know.”
What not to lead with
Your fee percentage. It appeared in none of the recommending quotes we found — the market's structural shift to contingency pricing means fee comparison isn't the citable unit here. Compliance credibility and technical specialism are what the engines in our sweep actually reached for.
Common questions
- Do I need new content, or can I update what exists?
- Start by auditing what exists against the two dated compliance changes (AIF, merged scheme) — fixing stale scheme framing is faster than new content and directly addresses what our sweep flags as a real, current risk.
- How do I know which engine matters most for my firm?
- Run the same buyer questions your firm actually gets asked across ChatGPT, Gemini and Claude separately, and compare recommendation rate, not just mentions. Our own sweep found identical, engine-specific behaviour independently in two unrelated markets, so there's a real pattern worth checking for your own.
- Does this apply if we also do general tax advisory work, not just R&D claims?
- The specific evidence in this page is drawn from R&D tax questions specifically — a firm working across broader tax advisory should treat each service line as its own measurement, since our wider research shows recommendation behaviour varies sharply by the exact question asked, not just by sector.
AI can't recommend what it doesn't know.
Start free. No commitment.