THE TOOLKIT
Appendix A — The Audit
The audit, run as one pass. Score every line pass, partial or fail across entity, reachability, corroboration, specificity, usability and measurement. Anything graded A that fails is urgent. Lines phrased as prohibitions invert: a fail there means you are doing something you should stop doing immediately.
From Becoming the Answer by Jeremy Osborn · 535 words
One pass. Score each line pass, partial or fail. Anything graded A that fails is urgent. Lines phrased as prohibitions mean the opposite: a fail there means you’re doing something you should stop.
Entity
| Check | Grade |
|---|---|
| A canonical facts page exists, with positioning sign-off | A− |
| Homepage or About page states identity in plain indexable text | A− |
| Organization schema present, with a complete sameAs | A− |
| Wikidata item exists, is referenced, and its Q-ID is in sameAs | A− |
| LinkedIn, Crunchbase and directory descriptions match the canonical statement | A− |
| Press boilerplate on the last ten releases matches current positioning | A− |
| Named experts have author pages, consistent bylines and author markup | B |
| Knowledge Panel claimed, if one exists | B |
| A quarterly consistency review has a named owner | A− |
Reachability
| Check | Grade |
|---|---|
| robots.txt does not disallow OAI-SearchBot, PerplexityBot or Claude-SearchBot | A |
| Nobody has added named crawler groups that bypass the global rules | A |
| Nobody has blocked GPTBot believing it controls citation | A |
| Nobody has blocked Google-Extended believing it controls AI Overviews | prohibition |
| Cloudflare and WAF settings audited against the September 2026 defaults | A |
| Key content appears in view-source, and in a curl fetch using each bot’s user agent | A |
| AI-crawler 404 rate measured, worst offenders fixed | A |
| No noindex or nosnippet on pages intended for AI answers | A |
| Bot mitigation reviewed against agentic browser traffic | B+ |
| Search Console AI opt-out not enabled | A |
Corroboration
| Check | Grade |
|---|---|
| Third-party roundups and review pages cited for your top ten commercial queries are mapped | B+ |
| Factual errors about you in those sources have been corrected | B+ |
| Digital PR targets the outlets that actually get cited | B+ |
| At least one original research study published or commissioned annually | B |
| A named, disclosed presence exists on the two key community platforms | B |
| A sentiment-neutral review solicitation program is running | B |
| Nothing has ever been seeded, bought, or sentiment-incentivized | prohibition |
| The FTC review rule is in every agency contract | A |
| A YouTube program exists with real substance | B− |
Specificity
| Check | Grade |
|---|---|
| A question inventory built from real customer language exists | B+ |
| Attribute intersections are mapped, not just head terms | B |
| Questions scored for winnability: current citations, current rank, information gain | B+ |
| Prices and specifications published wherever commercially possible | B+ |
| Constrained-recommendation pages exist for the top intersections | B |
| Fair comparison content exists, including against competitors | B |
| Content is dated and on a defined refresh cycle | B |
| No mass-generated permutation pages | prohibition |
Usability
| Check | Grade |
|---|---|
| Product feed complete, refreshed on a defined cadence | A |
| Product Highlight and Product Detail populated across the catalogue | A |
| Q&A attributes populated in the feed | A |
| Manuals and spec sheets attached via document links | A |
| /.well-known/ucp published, where Universal Cart is live | A− |
| Price, availability and fulfilment terms in the initial HTML | A |
| Quarterly agent-completability test run on core journeys | B+ |
Measurement
| Check | Grade |
|---|---|
| Search Console generative AI reports and Bing AI Performance enabled | A |
| GA4 AI Assistant channel verified, with the Direct gap documented | A |
| AI-bot log analysis running | A |
| A question set of 120–200, versioned and frozen quarterly | B+ |
| One synthetic-prompt tool and one real-prompt tool in place | B |
| Only two-to-four-week rolling averages reported | A |
| Mention rate and citation rate reported separately | A |
| Monthly accuracy audit of answers about the brand | A |
| Controllable inputs reported alongside outputs | B+ |
| Survey questions capturing AI-assisted research added | B |