Blog · Tech ·
How to show up in ChatGPT and AI-SEO in 2027
ChatGPT, Google’s AI Overviews, and the other answer boxes in 2027 do not share one ranking. ChatGPT search cites pages it can retrieve. Google’s AI features cite pages that are already eligible for Google Search. A tool that sells a score for “AI visibility” is usually selling a dashboard.
The work is the same shape as ordinary search, with two extra decisions: whether you allow the crawler that feeds citations, and whether the page says something a model cannot get from the ten sites that already said it. The foundation is how to rank on Google. The shared habits are SEO and GEO/AIEO best practice. This page is what changes when the result is a paragraph with a link under it.
Do not publish fifty versions of a question so a model might pick one. Google’s spam policies already treat that as scaled content abuse, including when the point was to manipulate an AI answer.
Buildex publishes this guide and builds the pages an answer engine would have to cite. The steps follow public documentation from OpenAI and Google. The build step is the one we sell, and this page says so. Being eligible is not the same as being quoted.
Where an AI citation comes from
Eligibility. A page that is not indexed, or that blocks the crawler an answer engine uses, cannot be the source. Training opt-out and search opt-out are different switches. People mix them up and then wonder why a block on one did nothing, or too much.
A quotable page. One question, a direct answer, then the detail, in text a crawler can read. A heading that matches how a person asks is useful. A page that only exists as a script, or as text baked into an image, gives the model nothing to lift.
Something that is yours. Google’s guidance for its own generative features is non-commodity content: a point of view, a first-hand detail, not a restatement of common knowledge. Answer engines already have the generic version. They cite the page that adds one.
The work, in order
In this order: say who you are, answer in the first screen, write what only you know, stay eligible in Google, decide the ChatGPT crawlers on purpose, ignore the special file, rebuild the page when the text is trapped, and check by asking the question.
Say who you are, the same way everywhere
An answer engine has to decide which “Buildex” or which clinic it means. Use one name, one description of what you do, and one address for the company across the homepage, the about page, and the profiles you actually control. Contradictions make you a guess.
Put that description in text. A logo is not an entity. The first sentence on the about page should be something a stranger could repeat.
Answer in the first screen
People ask complete questions. The page that gets cited often opens with the answer, then the conditions and the detail. A hero that says “Welcome to our innovative solutions” gives a model nothing to quote, so it quotes someone who was plainer.
One page, one question. A heading that is the question, a paragraph that is the answer, then the evidence. That is also how a human reads. You are not writing for a robot in a secret dialect.
Write the part only you know
A list of tips anyone could generate is commodity content. Google’s own guide for generative features on Search says a unique viewpoint and first-hand detail matter more than a special optimisation, and that restating common knowledge is the weak version. A method you use, a number from your work, a constraint you keep hitting: that is the sentence worth citing.
If you use a model to draft, the draft is not the page. Add the part that came from the work, and cut the part that could have been any company’s blog. Label nothing as a person who did not write it.
Stay eligible in Google
Google’s documentation is plain: to be eligible for AI Overviews or AI Mode, a page has to be indexed and eligible to show in Search with a snippet. There are no extra technical requirements, and no special markup that buys inclusion. Snippet controls that hide the text also hide you from the answer.
Crawling allowed, internal links, text on the page, structured data that matches the visible words. Those are the same fundamentals as ranking on Google. A page Google will not snippet is a page its AI features will not lean on.
Decide the ChatGPT crawlers on purpose
OpenAI runs separate bots. OAI-SearchBot is the one it uses to surface sites in ChatGPT search. OpenAI says a site that opts out of OAI-SearchBot will not be shown in ChatGPT search answers, though it can still appear as a navigational link, and that a robots.txt change can take about a day to apply. GPTBot is the training crawler. Disallowing GPTBot tells OpenAI not to use the content for training foundation models. The two settings are independent.
Allow OAI-SearchBot if you want to be eligible for ChatGPT search citations. Treat GPTBot as a separate decision about training, not as the switch for today’s answers. ChatGPT-User is a fetch a person triggered. OpenAI says it is not what decides search, and that robots.txt may not apply to it. Blocking “AI” in one line, without reading which bot, is how sites opt out of the thing they meant to join.
Skip the special file
Google’s guide on generative features says not to invent AI-specific files, special schema, or piles of near-duplicate pages in the hope of being chosen. A file that explains your site to a model is not a ranking system. Structured data helps when it describes the page a person can already see. It does not promote you into an answer.
The same warning covers the cottage industry of “AI SEO” checklists that tell you to add hidden text, fake authors, or a hundred question pages. That is spam with a new name. Write the page. Let it be fetched.
Rebuild when the words are trapped
A builder that paints the answer into an image, or hides it behind a widget that never lands in the HTML, cannot be quoted. Answer engines cite text. If the template will not give you a real heading, a real paragraph, and a stable URL, the template is the ceiling, the same way it is for ordinary search.
That is a custom website under website development: the words in the document, the identity consistent, the page fast enough that a person who followed the citation stays. You are not building a second site for robots.
Check by asking, then fix the page
Take five questions a buyer would ask. Ask them in ChatGPT with search, and in Google. See whether you are cited, whether a rival is, and whether the sentence they lifted is one you would stand behind. A missing citation with a blocked crawler is a robots.txt job. A missing citation with a vague page is an editing job.
There is no position three in an answer box. Some queries get no citation of you because a better page exists. Make yours the one worth quoting, keep it eligible, and read it again next month. Do not buy a guarantee.
What to do this month
This week, check eligibility
Confirm the important URLs are indexed and allowed to snippet. Then read robots.txt and decide OAI-SearchBot and GPTBot as two choices, not one.
This month, rewrite one page so it can be quoted
The question in the heading, the answer in the first paragraph, and one detail only you have. Ask that question in ChatGPT and in Google after it is indexed.
Ignore the file and the score
No special AI file, no mass of question pages, no dashboard that claims a percentage of “share of model”. Google has said the fundamentals are the requirement.
Rebuild when the text is not in the page
If the answer is trapped in a template, that is website development. The citation still depends on the words being worth taking.
When the pages worth citing need building, start with website development.
Common questions
If I block GPTBot, will ChatGPT stop citing me?
No. OpenAI’s documentation treats the crawlers as independent. GPTBot is training. OAI-SearchBot is what feeds ChatGPT search answers. You can allow search and disallow training. Blocking OAI-SearchBot is the change that keeps you out of those answers.
Does a special schema or an llms file get me into ChatGPT?
Google says generative features on Search have no extra technical requirement beyond being indexed with a snippet, and it warns against AI-specific files and schema tricks. ChatGPT search cites pages its search crawler can use. A file is not a citation. A clear page that is allowed to be fetched is the requirement.
Is AI-SEO different from SEO?
The eligibility is SEO: crawl, index, snippet, a page that answers. The extra part is which crawler you allow, and whether the first screen is quotable. Start with ranking on Google. The overlap is SEO and GEO/AIEO best practice. Come back here for the ChatGPT answer box. A site that does not rank and is not indexed will not become an AI source because the pitch said GEO.
