Full-Stack Marketing & Technology- Home
- AI Search Readiness Audit
Can AI Search Engines Understand Your Website?
An evidence-backed readiness audit for ChatGPT, Google AI Search, Perplexity and Claude — not an AI-ranking predictor. It crawls a sample of your pages and reports what is accessible, what is ambiguous, and what to fix, with every claim tied to what was actually fetched. No email required.
Starting
The four categories
What to do with this
This is what a crawler can see from outside in under a minute. If any of the above is worth a conversation, send us a brief.
Your Generative Engine Optimization Audit
Four transparent categories, scored from what was actually fetched.
Think of this as an AI SEO audit for how ChatGPT and Google AI Search read your site — not a conventional search-engine-optimization audit, and not a promise about rankings. "Generative engine optimization" (GEO) is the newer name for the same underlying discipline: making a site easy for a generative AI system, rather than a traditional search index, to find, parse and cite.
Every check is deterministic — a fixed rule applied to something the crawl actually returned, never a model's opinion. There is no AI-assisted assessment in this stage of the tool (that is a separate, later phase, kept clearly apart so nothing here is ever a guess dressed up as a measurement).
An unassessed category reports as unassessed, never as zero. Its weight leaves the score's denominator rather than dragging the number down — the difference between "we could not check this yet" and "this failed".
A site that blocks every crawler cannot score highly however well the rest of the site is built. Sensible caps apply to the headline number when a foundational check fails, and the report says exactly why.
| # | Category | Weight |
|---|---|---|
| 1 | Discoverable | 30 |
| 2 | Understandable | 25 |
| 3 | Structured | 25 |
| 4 | Citeable | 20 |
Which crawlers can reach your site?
Blocking a crawler is not automatically an error — which crawler matters. A search or answer engine's discovery crawler decides whether it can cite you at all; a training crawler only collects content to train a model, and refusing one is a legitimate, common publisher choice this audit reports without penalizing.
| Crawler | Owner | Purpose | What it does |
|---|---|---|---|
Googlebot | Search discovery / indexing | Crawls and indexes pages for Google Search. | |
Bingbot | Microsoft | Search discovery / indexing | Crawls and indexes pages for Bing, which also powers Copilot search. |
OAI-SearchBot | OpenAI | Search discovery / indexing | Crawls and indexes pages that ChatGPT search can cite. |
PerplexityBot | Perplexity | Search discovery / indexing | Crawls and indexes pages for Perplexity's own search index. |
Claude-SearchBot | Anthropic | Search discovery / indexing | Crawls and indexes pages that Claude's search feature can cite. |
ChatGPT-User | OpenAI | User-requested retrieval | Fetches one page when a person asks ChatGPT to open or read it. |
Perplexity-User | Perplexity | User-requested retrieval | Fetches one page when a person asks Perplexity to open or read it. |
Claude-User | Anthropic | User-requested retrieval | Fetches one page when a person asks Claude to open or read it. |
GPTBot | OpenAI | Model training | Collects content to train OpenAI's models. Blocking it has no effect on whether ChatGPT search can cite this site. |
ClaudeBot | Anthropic | Model training | Collects content to train Anthropic's models. Blocking it has no effect on whether Claude's search feature can cite this site. |
Google-Extended | Model training | Controls use of content to train Gemini and Vertex AI — separate from Googlebot, which crawls for Search. |
Can a crawler find every page on your site?
A sitemap is the list of pages a site is willing to hand a crawler directly, rather than making it guess from links alone. This audit reads robots.txt for a declared Sitemap: line, falls back to the common locations if there isn't one, follows a sitemap index into its child files, and reports which of it could actually be read — separating "no sitemap exists" from "the sitemap exists but errored," which are different problems with different fixes.
Indexability is the other half of the same question. A page can be in the sitemap and still tell every crawler to ignore it, via a noindex meta tag or an X-Robots-Tag header — sometimes left over from a staging build. That contradiction is worth catching on its own: listing a page and then disowning it sends a crawler two answers to the same question, and this audit checks for exactly that conflict rather than only checking the two facts separately.
Does your markup tell a machine who you actually are?
Structured data — JSON-LD, in schema.org's vocabulary — is the machine-readable layer sitting behind the words on a page. An AI system can, in principle, work out what a business does by reading prose; structured data tells it directly, in a shape built for parsing rather than for reading.
This audit checks for the schema types most directly tied to being understood and cited: an Organization or LocalBusiness entity naming the business, with sameAs links out to its other profiles — the detail that lets a system cross-check that the entity is real rather than merely asserted; FAQPage markup around real question-and-answer content, which labels the relationship between each question and answer; and BreadcrumbList, which states outright how a page sits inside the site rather than leaving that to be inferred from navigation.
None of this is scored as a functional requirement — a business with no FAQ content has nowhere to put FAQ schema, and that's a fact about the business, not a defect. What the report checks is whether the structured data that is there is present, valid, and pointed at the right things.
Readiness is not the same as being recommended
This tool measures readiness: can a crawler discover, access and parse your site. It cannot measure visibility — whether ChatGPT, Google AI Search or Perplexity actually cite or recommend your business for a given question. Visibility depends on the question asked, the model, your competitors, and live product behavior that changes over time and that no technical audit can observe from outside. A high readiness score is a floor a business needs to clear, not a guarantee of what sits above it.
No — and this audit treats it that way
llms.txt is a proposed, optional file some sites publish. Google Search's own documentation states plainly that it is not required or used for inclusion in generative search features. This audit checks for it — whether it exists, whether its links resolve — and gives it very little weight. Its absence is never a critical failure and never materially reduces the score.
What's actually in your control
Nobody outside OpenAI, Google or Anthropic can tell you whether a specific question will surface your business — that depends on the question, the model, your competitors, and product behavior that changes week to week. What is in your control is whether your site clears the technical floor those systems need before any of that can happen at all. In practice, most of that floor is the same handful of things:
- Don't block the crawlers you want to be found by. A wildcard
Disallow: /inrobots.txtremoves a site from every engine that honors it. Blocking a training-only bot like GPTBot or ClaudeBot is a legitimate choice and costs nothing here — blocking a search or user-fetch bot like OAI-SearchBot or Claude-SearchBot costs the citation itself. - Publish a sitemap and keep it accurate. A crawler that has to discover every page by following links finds fewer of them, slower, than one handed a list.
- Use structured data where it accurately describes visible content.FAQPage can label genuine question-and-answer content, but Google does not require special schema for generative search and no search or AI vendor confirms that the wrapper itself improves citation odds.
- Declare a real Organization entity, linked to your other profiles.A name and a
sameAslink or two is what lets a system confirm the business is real rather than merely claimed. - Name a real author and keep a visible "last updated" date current. An anonymous byline or a stale date both cost trust — the fix for a stale one is updating it, not removing it.
- Treat
llms.txtas optional, because it is. Fix everything above first — the measured evidence on how often AI platforms actually fetch it does not support spending real effort there before the basics are covered.
Every item above is something this audit actually checks — run it to see which of these your site already has, and which is the next one worth fixing.
Planning where to invest next?
Search visibility is only one acquisition decision. If paid media is also on the table, compare the economics with our Google Ads cost calculator,Facebook Ads cost calculator, orTikTok Ads cost calculator. See every free tool on the tools hub.
Common questions
It normalizes the address you give it, reads robots.txt, finds and reads the sitemap, crawls a representative sample of pages (the homepage, about/contact-shaped pages, and a spread of others, up to 10 in this stage), and checks whether named search and AI-search crawlers can reach the site, whether pages carry the structured data and metadata a search or answer engine reads, and whether the technical basics (titles, headings, canonical tags) are in place. Every check is deterministic — a fixed rule against what was actually fetched, never a guess or a model’s opinion.
Generative engine optimization is the practice of making a website easy for a generative AI system — ChatGPT, Google AI Search, Perplexity, Claude — to discover, understand and cite, the same way search engine optimization targets a traditional ranked results page. The mechanics overlap (a site still needs to be crawlable and well structured) but the target is different: a search engine ranks pages, a generative engine synthesizes an answer and decides which sources, if any, to name. This audit checks the GEO fundamentals — crawler access, structured data, and the other practices covered on this page.
A regular SEO audit is built around how traditional search engines rank pages — backlinks, keyword targeting, page speed, click-through rate. An AI SEO audit like this one is built around a narrower, more mechanical question: can a generative AI system's crawler actually reach the page and extract a clean, structured answer from it. Many of the same technical basics matter to both (a broken sitemap helps nobody), but this audit does not score keyword targeting, backlinks, or anything else a conventional SEO audit would — it is scoped specifically to AI-search and AI-citation readiness.
No, and this tool will never claim that. This measures technical readiness — whether a crawler CAN access, parse and cite the site. Whether an AI product actually recommends a business depends on the question asked, the model, competitors, and things no technical audit can see from outside. A readiness score is a floor, not a forecast.
No. It is optional and experimental, and Google Search does not require or use it for inclusion in generative search features. This audit checks for it and gives it very little weight in the score — its absence is never treated as a critical failure.
Not on its own. GPTBot trains OpenAI’s models; it has nothing to do with whether ChatGPT search can cite your site (that’s OAI-SearchBot). Blocking a training crawler is a legitimate, common choice, and this audit reports it without penalizing it. What does affect the score is blocking the crawlers that actually discover or fetch pages for a live answer — see the table below.
Because nothing else on the report matters if a crawler can’t get in. A site with perfect structured data and flawless page titles that also disallows every crawler from the entire site is not "mostly ready" — it is unreachable. The score reflects that rather than averaging it away against unrelated passing checks.
Start with what this audit checks, in order: make sure you aren't blocking the search or user-fetch crawlers (blocking training-only bots like GPTBot costs nothing), publish an accurate sitemap, use structured data where it truthfully describes visible content, declare an Organization entity with sameAs links to your other profiles, and keep a named author and a current "last updated" date on content that changes. See the "What's actually in your control" section above for the full list — none of it guarantees a citation, but all of it is a precondition for one.
No — different tool, different question. The Marketing Audit is a seven-axis brand and marketing health check with one page rendered and judged by a model. This audit crawls a sample of pages and checks discovery, structure and crawler access, with no model in the loop in this stage. They report separately on purpose.
Written by Daena Digital's engineering and marketing team. Last reviewed 28 August 2026. This audit measures technical readiness, not rankings, citations, traffic or revenue.
Want a person to look at what this can't see?
This audit checks technical readiness from outside. It cannot see your ad accounts, your first-party data, or whether your actual buyers are asking the questions you'd want to be cited for. If you want that read, tell us about it.