What every audit starts with: who the brand is, and what AI thinks it is
The first thing the audit captures is the brand as we’d describe it — positioning, category, market — so the rest of the document can measure the gap between that and what AI actually says. For this sample, the brand is us.
BeCited is a founder-led GEO audit service — flat-rate $2,000 for a one-week audit, $1,500/quarter for tracking. The category is contested by SaaS dashboards (Peec AI, AthenaHQ, Profound) on the tech side and SEO agencies adding GEO services on the consulting side. Our wedge is supposed to be price ($2k vs $10k+ engagements) and verification rigor — every quote read by hand, no fuzzy-matching. Whether AI engines know that — and recommend us for it — is the thing this audit measures.
What every audit produces: one score, four engines, the gap that matters
The score is a single number, but the story is almost never in the number — it’s in the gap between mentioned and recommended. For us, AI lists us 11% of the time and endorses us 8% of the time. A 4-point conversion gap with no source presence behind it — and that’s the diagnosis.
Engine-by-engine
Across 132 captures on ChatGPT, Perplexity, Claude, and Gemini, BeCited appears in 15 responses and is the primary recommendation in 10. That puts us 3rd in a market AI engines surface 103 distinct brands for — but the rank flatters the reality. Two of the top three (Peec AI, Profound) get cited 6–9× per audit-related query. We get cited once, sometimes twice.
The sharpest signal is the dimension breakdown: Source Presence at 5%. That’s the rubric saying “you are not on the platforms AI uses to answer this question.” Engine Consistency at 52% is the only above-floor number, and that’s because when any engine does find us, they roughly agree on what we are — the website signal is doing its job. The website isn’t the problem. The third-party ecosystem is.
Claude at 0% recommendation is the limit case: Claude leans on the widest, most editorial source set of the four engines, and it has nothing to cite. Until G2, Capterra, and a PCMag-tier placement exist, Claude has no path to recommending BeCited and the playbook below is what closes that.
What every audit produces: the ranked priority moves, in their actual order
The audit doesn’t output a wish list — it outputs 8 ranked priority moves, sorted by leverage (impact ÷ √effort × confidence). Seven carry high-impact labels. We publish the top five below, in the order the rubric returned them. Not curated. Not reordered.
The page above is the highlight reel. The audit goes deeper.
This sample shows the four blocks every audit produces: brand, diagnosis, playbook, and a closing view. The actual file you receive contains the same blocks plus a working layer underneath — the libraries, maps, tickets, and timelines a team can pick up and run with.
Above is the executive view. The full deliverable adds working artifacts — flip-target libraries with the exact competitor quote that beat you, per-page site readiness scoring, claim URLs with copy you can paste in, and a 90-day timeline. Everything is built from the same captured AI answers; this page just leads with the decision-grade summary.
- Flip target library. Side-by-side comparisons of the prompts where AI named you in passing but recommended someone else — with the exact winning competitor quote and a strategy to take that prompt back.
- Source ecosystem map. Every domain AI cited when answering questions about your category, tiered into primary / secondary / tertiary sources, with the unclaimed high-leverage platforms named.
- Site readiness audit. 18-check technical scan of how AI crawlers see your site — robots.txt, llms.txt, schema, quotability, information gain, render completeness, Core Web Vitals, plus per-page scoring on additional URLs you nominate.
- Per-engine job tickets. Each priority move expanded into a checklist with target URLs, exact copy you can paste into your CMS, success metrics, and the engine the move is meant to lift — ready to assign.
- Competitor interception plans. A per-competitor strategy for each of the top 5 threats — their positioning, their head-to-head record against you, and the specific signal you’d need to add to win their share.
- 90-day timeline + re-audit. Moves bucketed by when they realistically fit (this week / 30 / 60 / 90 days). Quarterly tracking clients get a re-audit with action attribution, score delta, and trend velocity.
Yes, we published our own F-grade. The rubric works because it grades us honestly too.
The standard pattern in GEO — and in SEO before it — is to publish only the audits that flatter the methodology. A SaaS dashboard that shows you “visibility” usually defines the term in whichever way makes its numbers look good, then publishes case studies of clients who scored well on that definition. We could have done that. We have a real client at 68/100 we could have led with.
BeCited’s rubric scored BeCited a 17 out of 100. We’re publishing it as the sample. If we don’t, the rubric doesn’t mean what we say it means — and a rubric that flatters its author isn’t measurement, it’s marketing.
The audit also returned what every honest audit returns when run on a brand-new SaaS: fix your third-party presence, surface your differentiators, earn editorial coverage. The same playbook a $30k agency would charge for. Different in our case only in that we’re running it on ourselves in public, with a re-audit date already on the calendar.
Three things this audit confirms about how the methodology behaves
- The rubric punishes brand newness. A six-month-old SaaS with no G2 listing, no PCMag review, and no Wikipedia entry will score in the teens regardless of website quality. Site readiness graded B; the GEO score still graded F. Site quality is necessary, not sufficient.
- The rubric separates “mentioned” from “recommended.” Visibility is 11%, recommendation 8% — a 4-point conversion gap. AI knows we exist; it has no language to endorse us over Peec AI or Profound. That gap is what positioning fixes, not visibility budget.
- Configured differentiators are testable. We declared six positioning claims at audit setup. The rubric measured whether AI surfaces each one. Five surfaced zero times. The diagnostic isn’t “your messaging is weak” — it’s “these specific six claims aren’t reaching the answer layer, here is which one to fix first.”