DEEN
Technical AI AuditOwned Intelligence

Can the AI even read your website? The technical audit tells you.

Before you measure whether AI systems fetch your content, the technical AI audit checks whether they can even read it. A guidance score from technical signals, per page, with history. Not a guarantee, but an honest baseline.

Run the free check See it in the Studio

Owned Intelligence · per page · history · guidance, not a guarantee

The problem

A page can look perfect in a browser and still be empty to the AI.

AI crawlers don’t open a browser. Whatever only appears via JavaScript, they don’t see. If the sitemap is stale or robots.txt blocks the wrong paths, the AI gets no page at all. If schema is missing, it gets a poorer one: readable, but with nothing telling it what is product, what is price and who the provider is.

Visible to humans

Fully rendered, nice, clickable.

In the browser everything loads, images, prices, FAQ. To a human visitor the page is complete.

Maybe empty to the AI

The crawler only reads the raw response.

No JavaScript, no waiting. What isn’t there server-side doesn’t exist for the bot. The technical audit makes exactly that gap visible.

What the audit checks

The technical signals that decide AI readability.

Server-side renderingIs the content in the raw response, without a browser executing JavaScript?
robots.txt & accessAre the relevant paths allowed for AI crawlers, and the right ones blocked?
llms.txtIs there a machine-readable signpost for which content AI systems should read first? Access itself is governed by robots.txt.
Schema.org / JSON-LDIs structured markup present, valid and server-side rendered?
SitemapComplete, current and reachable as an entry point for crawlers?
Status & redirectsClean status codes, no redirect chains, no error pages for bots?
Sitemap coverageHow many of the pages listed in the sitemap are actually reachable?
The score, three views

A value from 0 to 100 for the domain, plus the individual findings per page.

From foundations through solid and good to optimized. Every finding is marked critical, notice or OK, or as needing manual judgement or not measurable, and then it does not count toward the value.

Overview
The whole state at a glance
The score, the tier and the number of open findings for the whole domain.
Page explorer
A score per URL
So you see which pages lag, not just an average.
History
Development over time
Whether a fix really worked shows in the curve, not a gut feeling.

The audit is a guidance score from technical signals, not a guarantee of visibility in a specific AI service. It tells you whether the technical base is right, not whether ChatGPT recommends you today.

Readiness → Reach → Discovery

Readiness is the prerequisite. Reach is the proof.

The technical audit and Agentic Reach work together. Only once the base is right is it worth talking about reads and recommendation.

Technical AI Audit
Can the AI read your site?

The technical base: server-side, valid markup, clean access rules.

Agentic Reach
Does it actually read it?

Measured bot reads, by provider, touchpoint and product.
LLM Discovery
Do you show up in the answer?

Whether and where you’re named and recommended in AI answers.
Operation

Findings turn into concrete tasks.

  • Fix critical findings before anything else, that’s where real readability is lost.
  • Check before and after a launch or relaunch that AI readability holds.
  • Use the page explorer to improve the weakest URLs on purpose.
  • Use the history to prove a fix actually raised readiness.
Checked in concrete terms

Every check ends in a finding you can read back.

Every address is fetched with an AI crawler’s user agent and compared with what a browser shows after rendering. The Studio audit splits the result into eight areas with 19 individual signals and stores it per URL. This is what actually gets queried.

Reachability first

If the browser gets through and the bot doesn’t, that is not an accident.

The homepage is fetched twice, once as GPTBot, once as a regular browser. If the server answers the browser and blocks the bot with 403, 429 or a firewall challenge, that is a targeted bot block. If both are blocked, the report says inconclusive, not failed. If the domain is not reachable for bots, that stands as the most severe single finding at the top of the report, because every other finding depends on it.

Render gap in percent

Bots without JavaScript see a measurable share of your content.

The auditor compares the amount of text in the raw response with the rendered page and states the share in percent. Plus the list of what is missing: which main heading is absent, in its exact wording, how many prices are gone, which headings and how many words of body copy. Below 85 percent is a notice, below 50 percent a critical finding.

Reachability & WAFTwo fetches, bot against browser, plus the redirect chain hop by hop, up to five hops, with the status code at every stop.
Bot accessrobots.txt is parsed and evaluated separately for GPTBot, ClaudeBot, PerplexityBot and Google-Extended. In the Studio audit against the real content paths from the sitemap, not just against the root.
AI filesllms.txt and agents.md, each with a counter-check against HTML pages that answer with status 200 but are not the file.
Structured dataJSON-LD and microdata are read in and the @type values found are listed. It distinguishes between no markup, basic markup without Product or Service, and Product or Service markup present.
Structural consistencyAcross the crawled pages we check whether comparable content carries comparable structure: the same types, the same vocabulary, the same organisation block. Inconsistent markup is harder for a machine to read than none at all.
Rendering typeFrom framework markers, empty mount containers, visible text volume and script count, a page is classified as server-rendered, hybrid or built only in the browser. Detected frameworks include Next.js, Nuxt, Gatsby, Angular, SvelteKit, Vue and React.
Sitemap & metaSitemap reachable and machine-readable as urlset or sitemapindex, plus canonical and meta description. The list of individual pages comes from the sitemap, not from a link walk starting at the homepage.
Crawl efficiencyTime to first byte, response size, number of requests as well as status code and redirect target per page.
Language signalsWhich language a page declares and whether hreflang is set.

Not every signal carries the same weight. Reachability carries the most: if your domain is closed to bots, that stands as the most severe single finding at the top, however clean the rest may be. agents.md carries the least.

What deliberately stays out of the score: Core Web Vitals are collected and shown per page but do not count, they describe human perception, not bot readability. Google-Extended is not automatically treated as critical but flagged for manual judgement, because its effect cannot be derived cleanly.

New · ChatGPT Ads

And is your site ready for ChatGPT Ads?

Now that OpenAI serves ads inside ChatGPT, it is no longer only about whether AI can read your site, but whether it qualifies as an ad target. The audit checks three technical prerequisites that are measurable from the outside. As with the rest: an honest assessment, not a guarantee of approval.

OAI-AdsBot accessOpenAI’s ad crawler checks your landing page against the ad policy and uses the content for delivery. If your robots.txt blocks it, the ad will not be served, no matter how good the campaign is.
Conversion measurementWithout the OpenAI conversion pixel or the Conversions API there is no reliable attribution from click to outcome, and no outcome-based bidding. The audit detects whether measurement is present.
Structured product dataProduct and Offer markup helps above all with product-feed campaigns, where offer and landing page come from the feed, and makes your offer machine-unambiguous. OpenAI does not require it for ordinary landing pages; we treat it as a helpful signal.

What the audit deliberately does not do: judge your industry. ChatGPT Ads are blocked or heavily restricted for regulated fields such as gambling, alcohol and tobacco, weapons, financial and health topics, and political advertising, and self-service starts in the US first, with the EEA contextual-only at launch. Whether you fall into such a category is your call, not the checker’s.

Next step

Check in under a minute whether the AI can read you.

The free check shows you the technical state of your domain, no sign-up. The Studio audit goes deep, per page and over time.

Run the free check See it in the Studio

Owned Intelligence · guidance, not a guarantee · EU-hosted

Free check

Does AI even mention your brand?

Enter your domain. In about 60 seconds the check shows whether ChatGPT, Claude and Perplexity read and name you.

No sign-up. Free.