Is AI search optimisation worth it for an R&D tax adviser?
One firm in our sweep converted 62% of its AI mentions into a recommendation. A firm cited 36 times converted none. In a market this competitive for claim volume, that gap is the difference between a full pipeline and a firm nobody actually calls.
Steen Stones · Reviewed 5 Aug 2026
The honest answer depends on whether a firm can currently see the number that actually matters, and in this market almost none can. ForrestBrown, in our own sweep, converted 62% of its AI mentions into a specific recommendation. Another firm, named 36 times in the same sweep — more than most firms in the tally — converted zero. Both would look “visible” on a tool that only counts citations. Only one of them is winning claims through it.
Avouch multi-engine market sweep, 18 July 2026 — UK R&D tax credit advisers, ChatGPT + Gemini + Claude
Why this market rewards the work more than most
R&D tax claims are high-value and high-trust by nature — a business handing over its claim, and by extension its relationship with HMRC, wants a specific reason to pick one firm over another. Our sweep shows engines already answering that way: every recommendation we found cited a specific compliance process, technical specialism or track-record figure, never a general reputation statement. That's a market that rewards specific, evidenced positioning more than most, which means the return on doing this work properly is higher than in a market where any credible-sounding firm gets picked.
Where the return compounds
Across Avouch's wider research — 6,517 AI answers judged, every verdict stored and independently re-checked — comparison and ranking sites come out ahead of every other channel by a wide margin, however the analysis is read. For R&D tax specifically, our own sweep shows the equivalent signal is compliance-and-track-record content that reads as independently verifiable — an enquiry rate, a claim-value average, a specific process — not a directory listing in the traditional sense, but functioning the same way: a checkable, third-party-style credibility signal an engine can point to as the reason.
“There's a difference between showing up and being recommended. AI cannot recommend what it doesn't know.”
The three questions worth asking before you spend anything
- Do we know our mention-to-recommendation rate, per engine? Our own sweep found one major engine converting zero across two entirely separate markets — an average would hide exactly the progress that matters on the other two.
- Can we state a real, specific compliance figure — enquiry rate, claim-value average, sign-off process? Every recommendation quote in our data was built on a number or a named process, never a general trust claim.
- Is our content dated correctly against the current scheme? A page still describing the pre-merger SME/RDEC split is telling both buyers and engines the firm hasn't kept pace with HMRC's own changes.
Common questions
- How does this compare to the return on ordinary SEO for this sector?
- Ordinary SEO still matters — it's a real factor in whether an engine cites a page at all — but citation and recommendation are different outcomes in our data. A firm can rank well and still be the one converting zero of its mentions, exactly like the 36-mention, zero-recommendation firm in our own sweep.
- Is claim value the only thing that matters for conversion?
- No — sector specialism (biotech, construction, mid-market) and stated compliance process both did real work in the quotes we found. Claim value is one strong signal among several, not the only lever.
- What's a realistic first step?
- Establish the baseline most firms have never measured: how often is the firm named across the questions its buyers ask, per engine, and how many of those mentions become a specific recommendation. Everything else in this territory follows from knowing that number.
AI can't recommend what it doesn't know.
Start free. No commitment.