What the search providers themselves say
Google is the only large provider with a detailed public page on the subject. In AI features and your website it states that there are no additional requirements to appear in AI Overviews or AI Mode: a page has to be indexed and eligible to be shown in Search with a snippet. No new machine-readable files and no special schema.org markup are needed. The practices it lists are ordinary Search fundamentals:
- Let Googlebot crawl the page, including through your CDN or the hosting firewall.
- Link to important pages internally so they can be found.
- Keep important content available as text, not only inside images or video.
- Make sure structured data matches the visible text.
- Offer a good page experience, and keep Business Profile and Merchant Center information current.
The usual preview controls apply: nosnippet, data-nosnippet and max-snippet limit what can be shown, and noindex removes the page from Search. Google-Extended is a separate control for AI training and grounding in some of Google's other systems, not for Search.
OpenAI documents one rule that matters just as much: sites that block its search crawler OAI-SearchBot will not be shown in ChatGPT search answers, while blocking GPTBot only opts out of model training (OpenAI crawler documentation). Beyond crawler access, no provider publishes how it picks sources, so any claim about what "AI prefers" is a guess unless it comes with data.
What the Sitequiry GEO score checks
The GEO score starts at 100 and loses points for each missing signal. Every signal is read from the HTML of the one page you submit, so in the report they carry the label "detected". None of them is a measurement of how an AI system behaves.
| Signal | What counts | Points if missing |
|---|---|---|
| Amount of text | Words of visible text. Under 150 is flagged as thin content; 150 to 399 gets a smaller deduction. | 25 or 10 |
| Heading structure | Exactly one H1 and at least one H2. Otherwise 12 points go when there is no H1 and 6 in any other case, plus 8 more when there is no H2. | Up to 20 |
| Who is behind the page | Organization-type structured data, a company name with a legal form (for example d.o.o., Ltd, GmbH, SARL, S.r.l. or S.L.) or the site name taken from the domain. Not checked on articles. | 15 |
| A place | A short sentence with a place or address word. Not checked on articles and product pages. | 5 |
| Main topic | A heading, title or meta description of 12 to 90 characters that is not boilerplate. Not deducted on articles. | 8 |
| Summary | A meta description, plus at least 120 characters of body text. | 8 |
| Byline and date (articles only) | A visible author or newsroom credit, and a publish or update date. | 12 and 10 |
Two checks are lenient by design, and you should know which. The site name from the domain counts as an identified source, so "who is behind the page" almost never fails: a pass means the minimum is there, not that the page explains convincingly who you are. The place check matches a short list of words, such as Ljubljana, Maribor, Vienna, "office" or "location", so a town outside that list goes unnoticed. Read the score as a checklist of basics, then judge the copy yourself.
What no tool can measure from the outside
Whether ChatGPT, Gemini, Perplexity or Google AI Overviews cite your page, for which questions and how often, is not visible to Sitequiry or to any other outside crawler. Answers vary with the wording of the question, the user and the day, and the providers do not publish citation logs.
Some traces do exist. Google counts clicks and impressions from AI Overviews and AI Mode in Search Console, inside the "Web" search type of the Performance report, together with ordinary results. Referral visits from AI services in your own analytics show that someone followed a link from them. Asking an assistant about your business by hand is a spot check, not a measurement.
Before and after: a summary that can be quoted
The following is an illustration with an invented business, not a real case. Many service pages open like this:
Before: "Welcome! We are a team of passionate professionals offering high-quality solutions tailored to your needs. Contact us to find out more."
After: "Brzica is a bicycle repair shop in Maribor. We service city, road and electric bikes, including Bosch and Shimano motors. A standard service costs from €45 and usually takes two working days. Open Monday to Friday, 8:00 to 17:00."
The first version could describe any company in any country, so there is nothing in it worth quoting. The second says who, what, where, how much and how long, and each sentence still makes sense when it is lifted out on its own. The name, the price and the hours are made up for this example; on a real page every figure has to be true and kept up to date.
GEO and AEO: where the line runs
Sitequiry scores the two separately. AEO covers access and machine-readable signals: which AI crawlers robots.txt lets in, whether an llms.txt file exists and what structured data the page carries. GEO covers the content those crawlers find once they are in. A page can pass one and fail the other: an open robots.txt does not help a page that says nothing specific, and a precise page is of no use to ChatGPT search if OAI-SearchBot is blocked. The free AI crawler checker answers the access question for a whole domain in seconds.
What our study of 90 websites found
At the end of September 2026 Sitequiry crawled up to 20 pages on each of 106 websites for its AI search readiness study; 90 crawls completed. Two results relate directly to the GEO checks:
- 30 of 85 analysable homepages (35%) carried Organization markup or a subtype such as LocalBusiness.
- On 35 of 81 sites (43%), two or more indexable pages had no H1 heading.
It is a convenience sample, not representative of the web, and the study did not measure whether any of these sites is cited in AI answers.
Where to start
- Check access first: Googlebot and the AI search crawlers you want must not be blocked, neither in robots.txt nor by bot protection at your CDN or host.
- Rewrite the opening paragraph of your most important pages so it states who, what and where, and price and timing where they apply.
- Give each page one H1 and H2 subheadings that match the questions customers actually ask.
- On articles, show the author and the date of the last real update.
- Keep the facts on the page, in your structured data and in your Google Business Profile consistent.
Google AI Overviews and AI Mode covers the controls and how to measure the effect. To see where your own page stands, run the free analysis.
Frequently asked questions
Is GEO different from SEO?
Less than the name suggests. For Google, the documented requirements for AI features are the same as for Search. The difference is emphasis: Google asks for important content to be available as text, and a system that writes answers can only reuse what it can read on the page.
Can anyone promise that ChatGPT or Google will cite my site?
No. The providers decide which sources to use and do not publish how. Treat any promise of AI citations as a warning sign.
Do I need llms.txt or special markup for GEO?
Not for Google: its AI features need no new machine-readable files and no special schema.org markup. llms.txt is an informal proposal; Sitequiry reports whether you have one under AEO, and no major provider has committed to using it.
Why is my GEO score high when my page never shows up in AI answers?
Because the score checks basics on your own page, not how it compares with the other sources an AI system could use. Competing pages, the wording of the question and the provider's own choices all matter, and none of them is visible in your HTML.
Check the GEO signals on your page
One URL, free, with SEO, speed and AEO in the same report.