Example of a full report
11-page PDF, no email required — console guide and marketing plays included — score, trajectory, platform radar, entity graph, citability and the ordered plan.
Score trajectory
Each recommended fix added in the order the engine ranks it — not an estimate, the audited point value of every check.
Platform radar
The five engines, as a shape.
Where the points are
Eight categories. Green ≥80 %, amber ≥60 %, red below.
Citability
Twelve methods, each scored 0–3. Half the zeroes are structural, not editorial.
Citability 53/100. The four zeroes — quotations, outbound citations, authorship, freshness — are the fastest to fix and the ones assistants quote most.
Band ladder
Where 71 sits today, and where the fixes land it.
What the audit found
A French DPE-verification tool with a clear offer and solid on-page basics. The losses are concentrated in entity signals, AI discovery and llms.txt depth — not in the page copy.
| Category | Points | Finding |
|---|---|---|
| robots.txt | 15/18 | All 27 AI crawlers allowed — but only through User-agent: *; the five citation bots have no explicit rule. |
| llms.txt | 12/18 | File present with 20 links, all resolving against the sitemap. Too thin for depth points, no llms-full.txt. |
| Schema JSON-LD | 13/16 | Organization, WebSite, WebApplication and FAQPage — no Article, so editorial pages cannot carry dates or authors. |
| Meta tags | 14/14 | Title, description, canonical and Open Graph complete. |
| Content | 11/12 | 906 words, H2/H3 structure and lists. No outbound citation. |
| Brand & entity | 3/10 | Name differs across title, H1 and schema; no sameAs; no about link; schema and meta descriptions disagree. |
| Signals | 3/6 | lang present; no RSS link; no dateModified. |
| AI discovery | 0/6 | /.well-known/ai.txt, /ai/summary.json, /ai/faq.json and /ai/service.json all absent. |
Platform readiness
The same evidence, mapped to the five engines buyers ask about. A platform score weighs only the signals that engine actually uses — it is not a second opinion on the overall score.
| Platform | Score | Band | Main gap |
|---|---|---|---|
| Google AI Overviews | 100 | ready | No blocking gap: Googlebot allowed, Organization + FAQ, extractable structure. |
| ChatGPT | 100 | ready | No blocking gap: OAI-SearchBot allowed, llms.txt present, FAQ page. |
| Perplexity | 65 | partial | No outbound citations; no freshness date. |
| Gemini | 60 | partial | No Knowledge Graph links; brand name inconsistent across sources. |
| Microsoft Copilot | 80 | ready | Bingbot allowed and schema present; no dateModified. |
Scores derive from robots rules, llms.txt, schema types, structure and freshness — the signals each engine relies on. They do not claim the engine cites the page.
Entity graph — what an engine can resolve
Four schema nodes exist, three are missing, and no sameAs link anchors the brand to the Knowledge Graph. Green is present, red is missing, dashed edges are the links that do not exist yet.
An assistant that cannot resolve the entity quotes the idea and drops the brand. sameAs plus Article schema is the cheapest seven points in this audit.
How to read the console
The same payload the console renders, explained once. Every number points back to a check in this report.
- The KPI row — four questions: how readable (GEO), how quotable (citability), how fast to update (IndexNow), how many engines are ready.
- The trajectory — the recommended fixes added in the engine's order, so point 6 is the cheap half and point 16 the full plan.
- The engine ranking — the gap between each engine and 100. A long red delta means the gap is structural, not cosmetic.
- The strip and the tabs — the eight categories ranked by cost to fix, then Search, IndexNow and Tracking at full size.
Start with Overview, then GEO. The other three tabs answer the follow-up questions: demand, speed, and whether the measurement itself can be trusted.
Three plays this audit unlocks
Each play is anchored to a check below it — no revenue projection, just the points it moves and the signal to watch.
1. Be the brand, not the page +7 GEO · +3 citability
Entity signals sit at 3/10: sameAs has 0/4 pillars, the name differs between title, H1 and Organization schema, and there is no about link.
Do: one name everywhere, sameAs to Wikidata, LinkedIn and Crunchbase, a real about page, a Person author with a URL.
Watch: branded queries in Search Console, and whether an assistant repeats “OpenDPE” when asked who verifies a DPE.
2. Publish the answer, not the brochure +12 citability
Citability is 53/100: 0/3 on quotations, outbound citations, authorship and freshness. The page states numbers but sources none of them.
Do: quote the official DPE method with a link, turn one comparison into a table, name an author, add dateModified, and put a sourced figure above the fold.
Watch: the four zeroed methods, re-run monthly, then AI referral sessions in GA4.
3. Own the launch window +45 IndexNow · +6 discovery
IndexNow is 35/100 — no key file — and AI discovery is 0/6: no /.well-known/ai.txt, no /ai/*.json. New pages wait for a crawl nobody announced.
Do: publish the key file, verify it, announce each release in one POST, and ship the machine-readable summary, FAQ and service files.
Watch: time between publish and first crawl in Bing Webmaster Tools; AI referral sessions after a launch.
Citability — would the passage be quoted?
12 methods, each scored 0–3. Green is full credit, amber partial, red missing. Evidence is taken from the page itself.
Citability 53/100. The four zeroes — quotations, outbound citations, authorship, freshness — are the fastest to fix and the most quoted by assistants.
Critical findings
1. The entity layer is the weakest category: 3/10
An AI engine has to decide that “OpenDPE”, the title, the H1 and the Organization record are the same thing. Today they are not aligned.
name: “DPE d'un bien a vendre…” (title/H1) vs Organization schema · sameAs: 0/4 · about link: none
Fix: one brand name across title, og:title, H1 and schema; add sameAs to LinkedIn/Wikidata; link a real about page; reuse the meta description as the Organization description.
2. No AI discovery endpoints: 0/6
None of the four discovery files exist, so an agent has no machine-readable summary, FAQ or capability statement to route on.
GET /.well-known/ai.txt → 404 · /ai/summary.json → 404 · /ai/faq.json → 404 · /ai/service.json → 404
Fix: publish the four files (the generator emits valid shapes); they are static JSON and take minutes.
3. llms.txt is valid but thin: 12/18
Twenty links all resolve, so drift is clean — but the file carries fewer than 1,000 words and has no expanded companion.
20 links · 100% resolve · 11 sitemap pages never linked · no /llms-full.txt
Fix: generate with the sitemap, then add product, documentation and company sections; ship /llms-full.txt for depth.
4. Crawler access is one wildcard edit away
All 27 bots are allowed — via *. A single change to that block would silently remove every AI crawler, and the citation bots have no rule of their own.
robots.txt: 1 group (`*`), 0 explicit citation-bot rules · crawl-delay: none
Fix: add explicit Allow groups for OAI-SearchBot, Claude-SearchBot, PerplexityBot, Googlebot and Applebot.
5. Citability 53/100: the copy is not quotable yet
No quotations, no lists or tables in the inspected page, no question-shaped headings, no outbound citations, no author and no freshness date.
methods passed: 5/12 · weakest: quotations, structure, question headings, citations, authorship, freshness
Fix: add a sourced pull-quote, convert one comparison into a table, phrase two H2s as questions, cite three sources, add an author and dateModified.
Action plan
Ordered by the server's own logic: priority bucket, then recoverable points. Points are exact — they are the difference between what a check awards today and its maximum.
| # | Priority | Check | +Points | Action |
|---|---|---|---|---|
| 1 | high | robots.citation_bots | 3 | Explicit Allow rules for the five citation bots. |
| 2 | medium | schema.article | 3 | Add Article markup with author and dates. |
| 3 | medium | brand.kg_readiness | 3 | Add sameAs to Wikipedia, Wikidata, LinkedIn, Crunchbase. |
| 4 | medium | llms.depth | 2 | Expand llms.txt past 1,000 words with real sections. |
| 5 | medium | llms.depth_high | 2 | Ship /llms-full.txt for the 5,000-word tier. |
| 6 | medium | llms.full | 2 | Link the full reference from llms.txt. |
| 7 | medium | brand.name_coherence | 2 | One brand name in title, og:title, H1 and schema. |
| 8 | medium | brand.desc_coherence | 1 | Reuse the meta description as the schema description. |
| 9 | medium | brand.about_link | 1 | Link a visible about/company page. |
| 10 | low | signals.rss | 2 | Link an RSS/Atom feed in the head. |
Top six fixes → 86/100. Every failing check → 100/100 (29 points, capped). Scores are deterministic: re-running the same snapshot reproduces them exactly.