SEO SEO-05

SEO in the AI Era: What to Do for Google and ChatGPT

Google says there are no special optimizations for its AI features, and OpenAI splits its crawlers by purpose. A five-step operating plan, grounded in official documentation, for deciding what to open up, what to control and what to improve.

Published
Read 13 min
Table of contents

Most companies asking “how do I do SEO for AI?” expect a list of new tactics. The official documentation says nearly the opposite: Google states that there are no additional requirements or special optimizations to appear in AI Overviews or AI Mode, and that long-standing SEO best practices remain relevant. OpenAI, for its part, does not describe content tactics at all. It describes crawlers and how to control them in robots.txt.

That is the thesis of this article. The real work is not inventing a new discipline. It is five specific decisions: which crawlers you allow, whether your pages are eligible, which preview controls you apply, whether your content offers something a summary can’t replace, and how you measure without fooling yourself. None of them guarantees a place in a generative answer. All of them raise the odds that your content is eligible and remove the mistakes that quietly exclude you.

This is an operating plan. If you first need to understand what SEO, AEO and GEO mean and how they relate, start with SEO, AEO, and GEO: The Visibility System. We assume that foundation here. One scope note: “SEO with AI” can mean using AI tools to do SEO work, or preparing your site to be visible in Google’s AI search and in ChatGPT. This article covers only the second.


What the Primary Sources Say (and Don’t Say)

It helps to separate what is documented from what the market repeats. This is what the primary sources establish, as of the date we consulted them (October 9, 2026):

Verifiable statements in the official documentation and what they mean in practice
Source What it establishes What it means for your business
Google, AI features and your website To be a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to be shown in Search with a snippet. There are no additional technical requirements or special optimizations. Your foundation is standard Search eligibility. If that fails, no "AI" tactic makes up for it.
Google, AI features and your website You don't need new machine-readable files, AI text files or new markup, and there is no special schema.org type. Don't spend budget on files or markup "for AI" expecting to enter Google's features.
Google, AI features and your website The controls are nosnippet, data-nosnippet, max-snippet and noindex; Googlebot's robots.txt controls crawling for Search. Google-Extended is a separate control for training and grounding in other products. You have control levers, but each one carries a visibility cost.
OpenAI, Overview of OpenAI crawlers OAI-SearchBot surfaces sites in ChatGPT's search features; GPTBot supports model training; ChatGPT-User handles user-initiated actions. Each setting is independent. You can allow search and decline training. They are separate decisions.
Google, Top ways to ensure your content performs well (May 2025) Recommends unique, non-commodity content, a good page experience, accessible content, preview controls, and structured data that matches visible text. It's the same SEO as before, with the emphasis on what can't be summarized without loss.

Sources: Google Search Central (2025 and later updates); OpenAI Platform (crawler documentation). Consulted October 9, 2026. These pages change; confirm the current version before deciding.

Two caveats in the interest of honesty. First, these sources describe eligibility conditions, not selection mechanisms. None of them explains why a particular link appears in a particular answer, and anyone who explains it in detail is interpreting, not quoting. Second, Google states that meeting every requirement does not guarantee it will crawl, index or serve your content.

With that boundary clear, here is the plan.


Step 1: Set Your Access Policy Before You Touch the Content

The first decision isn’t editorial. It’s a business decision: what do you want AI systems to do with your content? There are three distinct uses, and the documentation separates them:

  1. Appearing in answers with a link (search). This benefits anyone who wants to be discovered.
  2. Training models. You may want to allow it, or not.
  3. User-initiated actions (for example, someone asking ChatGPT to read your page). OpenAI notes that because these actions are initiated by a user, robots.txt rules may not apply.

The table summarizes the relevant crawlers and tokens, and what each controls according to official documentation:

What each crawler or robots.txt token controls, according to official documentation
Crawler or token What it controls If you block it
Googlebot Crawling for Google Search, which includes its AI features (Google says AI is built into Search) You leave Search and its AI features
Google-Extended Whether content Google crawls may be used to train future Gemini models and for grounding in Gemini Apps and Vertex AI Per Google, it does not affect inclusion in Search and is not a ranking signal
OAI-SearchBot Surfacing sites in ChatGPT's search features You reduce or remove your presence in those answers
GPTBot Use of content to train OpenAI's foundation models You signal that your content should not be used for training
ChatGPT-User User-initiated actions in ChatGPT and custom GPTs OpenAI warns robots.txt rules may not apply

Sources: Google Search Central, AI features and your website and Google's common crawlers; OpenAI Platform, Overview of OpenAI crawlers. Consulted October 9, 2026.

Here is one example of a possible policy, not a universal recommendation. A services firm that wants to be discovered in ChatGPT but prefers not to hand its content library over for training could allow OAI-SearchBot and disallow GPTBot and Google-Extended.

User-agent: OAI-SearchBot
Allow: /

User-agent: GPTBot
Disallow: /

User-agent: Google-Extended
Disallow: /

Three operational notes that prevent common mistakes:

  • Check the CDN and firewall too. Google’s best practices include making sure crawling is allowed “in robots.txt, and by any CDN or hosting infrastructure.” A correct robots.txt does nothing if a security rule returns errors to crawlers.
  • Wait before concluding anything. OpenAI says robots.txt changes can take about 24 hours to be reflected in its systems.
  • Blocking is not always absolute. OpenAI’s publisher FAQ says that if it obtains the URL of a disallowed page from a third-party search provider or by crawling other pages, and sees signals the page is relevant, it may show just the link and page title in ChatGPT Atlas. Its stated way to prevent that is a noindex meta tag, which a crawler can only read if the page is allowed to be crawled.
  • Verify crawlers; don’t trust the user agent string alone. OpenAI publishes its crawler IP ranges as JSON files. Your technical team can use them to validate real traffic in server logs.

Make this call with someone who answers for the business, not only the dev team. It is an intellectual property decision and a visibility decision at the same time.


Step 2: Confirm Google Can Show You With a Snippet

Google’s condition is literal: indexed and eligible to be shown with a snippet. That changes the order of your audit. Before talking about content for AI, answer these questions with Search Console data:

  • Are the pages that matter indexed? (the page indexing report)
  • Do they return a 200 status code and indexable content? Google lists both explicitly among the requirements for being considered.
  • Are noindex, nosnippet or max-snippet directives applied by a template, plugin or legacy setting? This can happen on sites that have been through several migrations or theme changes.
  • Is the important content in text, not only inside images or elements that don’t render?

The tool for checking what Google actually sees is Search Console’s URL Inspection, which shows the HTML Googlebot received. It’s the same tool Google recommends for debugging preview controls.

The outcome of this step is usually unglamorous and valuable: a short list of important pages that aren’t eligible because of a configuration error. Fixing that has a better expected return than any “AI rewrite.”


Step 3: Use Preview Controls Deliberately, Not Out of Fear

The controls exist and they work for AI features: Google confirms that nosnippet, data-nosnippet, max-snippet and noindex apply, and that more restrictive permissions limit how your content is featured in its AI experiences.

The useful question is what you are protecting and at what price. Three scenarios:

  • Commercial content you want discovered (services, guides, comparisons): there is usually no reason to restrict it. Limiting it also shrinks your storefront in classic results.
  • Paid or high-value proprietary content (reports, data, client material): it may make sense to limit snippets or gate access, accepting that you’ll lose some visibility.
  • Specific fragments within a page (a pricing table, a sensitive figure): data-nosnippet lets you protect one block without removing the whole page.

One practical detail Google flags: a change isn’t instant. Google says crawling can take anywhere from several days to several months depending on how often its systems decide a page needs refreshing, and that you can request a recrawl.


Step 4: Content That Survives a Summary

This is the only truly editorial work. Google puts it this way in its guidance for site owners: create unique, non-commodity content that visitors find helpful and satisfying, in an environment where users ask longer, more specific questions and follow-up questions.

As a decision criterion, a piece of content earns its place if it passes this test: if a good three-line summary can replace it without loss, there is no reason for anyone to click, and no reason for a system to prefer it over ten equivalent sources. What tends to survive a summary:

  • Original data or analysis, with the methodology explained and dated.
  • First-hand experience and judgment: which decision we recommend, under what conditions and with what exceptions.
  • Tools, templates and processes the reader uses, not just reads.
  • Verifiable information about who wrote it and how. Google’s helpful content guidance proposes the “Who, How and Why” questions: clear authorship, transparency about automation, and a purpose of helping people rather than manipulating rankings. We go deeper on building that credibility in E-E-A-T for Mid-Sized Companies.

And a warning about the other side. Google says using generative AI to produce many pages without adding value for users may violate its policy on scaled content abuse. AI can help structure and research; judgment, fact-checking and a point of view remain human. We lay out the boundary in Generative AI in Content Production.

One point for anyone worried that long questions will break their keyword strategy. Google explains that AI Overviews and AI Mode may use a “query fan-out” technique, issuing multiple related searches across subtopics to build a response. The practical consequence is not to write hundreds of long-tail pages. It’s that covering a topic in depth with a coherent structure, a site whose pages reinforce each other, makes more sense than spreading thin. That is the argument behind content architecture for visibility and topical authority.


Step 5: Structure, Structured Data and Page Experience

The three technical elements Google names explicitly are modest:

  • Important content in text. If the key answer exists only in an image or a PDF without text, it isn’t accessible to either logic.
  • Structured data that matches what’s visible. Google asks that everything in your markup also be visible on the page and that you validate it. There is no special schema for AI. Useful markup is still whatever fits the page type (organization, article, product, FAQ where appropriate).
  • A good page experience, for people arriving from classic or AI results alike: it displays well across devices, loads quickly, and lets visitors tell the main content from everything else. The link between technical performance and conversion is covered in Technical SEO and User Experience.

Google adds two recommendations that rarely show up in “SEO for AI” guides: support text with high-quality images and video where relevant, and keep Merchant Center and Business Profile information up to date. For local businesses and retail, those listings are part of the storefront.


What You Don’t Need to Do (and Often Get Sold)

Based on what the documentation says and doesn’t say, these practices are often presented as mandatory without official backing for Google’s features:

  • Creating an “AI” text file or special schema to enter AI Overviews. Google says it isn’t necessary. If you decide to publish a file like that for other reasons, treat it as a low-cost bet, not a proven lever.
  • Rewriting the whole site in “answer format.” Structuring content well is good SEO. Chopping it up until it loses its argument turns it into the commodity Google advises against.
  • Changing dates to look fresh. Google’s helpful content guidance lists changing page dates without changing the content among the signs of content made for search engines rather than people.
  • Paying for guaranteed “ChatGPT ranking.” No official documentation allows anyone to promise it. For this specific topic, what works and what is hype when appearing in ChatGPT details what evidence exists.
  • Blocking everything as a precaution. Walking away from OAI-SearchBot or Googlebot out of fear of AI means giving up visibility to avoid a risk you can manage with finer controls.

How to Measure Without Fooling Yourself

Two documented facts set the frame. First, Google includes AI Overviews and AI Mode traffic in the Search Console Performance report, under the Web search type; a click on an external link in an AI response counts as a click. Second, Google launched a dedicated generative AI performance report (announced June 3, 2026 and rolled out to all sites on August 31, 2026) showing impressions, pages, countries, devices and dates. The report’s documentation lists no clicks or queries, and those impressions are already counted in the main report.

That means Search Console tells you how often you appear, not how much influence that appearance has. For ChatGPT, OpenAI documents that publishers can track referral traffic with analytics tools. The details and limits of that measurement, along with Bing and GA4, are covered in how to measure brand visibility in AI answers.

One last point of judgment. Google maintains that clicks from pages with AI Overviews are higher quality (users spend more time on the site) and recommends looking at the overall value of visits, using conversion indicators like sales, sign-ups or inquiries, not just clicks. That is Google’s claim about its own data. Treat it as a hypothesis you can test with your analytics, not as a fact about your business. For the other side of the equation, the effect of direct answers on click volume, see zero-click search: how to build visibility.


A Ten-Point Checklist

Editorial framework · Maccam Network

  1. Access: decide and document your policy

    Write down which crawlers you allow (Googlebot, OAI-SearchBot) and which you decline (for example GPTBot or Google-Extended), with a business owner named. Review robots.txt, CDN and firewall together.

  2. Eligibility: indexing and snippets on key pages

    For your 20 most important commercial pages, check indexing, 200 status, no unintended noindex or nosnippet, and content in text. Fix this before producing new content.

  3. Controls: restrict only with an explicit reason

    If you use nosnippet, data-nosnippet or max-snippet, note what you protect and what visibility you accept losing. Review quarterly for inherited restrictions.

  4. Content: pass the summary test

    For each priority page, ask what it offers that a three-line summary can't replace: proprietary data, judgment, a tool or experience. If there's no answer, it's a candidate for rewriting or merging.

  5. Authorship and transparency

    Identifiable authors, real dates, cited sources, and a clear note on how content was produced when automation was involved.

  6. Coherent topic coverage

    Group pages by topic with internal links between them. A well-covered topic answers chained questions better than isolated pages.

  7. Structured data faithful to what's visible

    Markup appropriate to the page type, validated, with nothing that doesn't appear on the page. Don't hunt for an "AI schema."

  8. External listings up to date

    Google Business Profile and, if you sell products, Merchant Center, with details consistent with your website.

  9. Measurement with declared limits

    Search Console (performance and generative AI reports), analytics with AI referral segments, and a record of what each source does not measure.

  10. Honest expectations

    No step guarantees appearing in an answer. The goal is to be eligible, useful and verifiable, and to measure whether that turns into visibility and qualified inquiries.

The ten checks derive from the official documentation cited below and from our working method. They are a work order, not a promise of results.


Next Step

If your team isn’t sure whether the pages that carry the business are actually eligible, which crawlers to allow, or how to measure the effect without claiming credit that can’t be proven, the starting point is a scoped technical and editorial review. It’s part of our strategic SEO service and builds on Maccam Network’s SEO methodology. If you’d rather talk it through first, get in touch.


Sources

Sources verified as of October 9, 2026.

Preguntas frecuentes

According to Google's documentation on AI features, no. It states there are no additional requirements or special optimizations, and that SEO best practices remain relevant. To appear as a supporting link, a page must be indexed and eligible to be shown in Search with a snippet. Meeting the requirements does not guarantee that Google will crawl, index or serve the content.

That depends on your policy for using your content to train models. According to OpenAI, GPTBot is used to improve its foundation models, and disallowing it signals that your content should not be used for training. OAI-SearchBot is a separate crawler used to surface sites in ChatGPT's search features, and each setting is independent of the others. Blocking GPTBot does not remove you from ChatGPT search, but blocking OAI-SearchBot limits your presence there. OpenAI says robots.txt changes can take about 24 hours to take effect.

Not for Google's AI features. Its documentation says you do not need to create new machine-readable files, AI text files or markup, and that there is no special schema.org structured data to add. It does ask that any structured data you use match the visible text on the page. For other systems, check their own documentation: none of the sources we reviewed guarantees results from adding these files.

Google lists nosnippet, data-nosnippet, max-snippet and noindex as the controls, and says more restrictive permissions limit how your content is featured in its AI experiences. Crawl access for Search is still controlled with robots.txt rules for Googlebot. Weigh the cost before applying them: a more restricted snippet is also a smaller storefront in classic results.

No. Neither company's official documentation offers that guarantee, and Google states explicitly that indexing and serving are not guaranteed even when every requirement is met. What you control are the eligibility conditions: crawling is allowed, the page can be shown with a snippet, and the content is original and useful. Be wary of any vendor that promises placement in generative answers.

Traffic from AI Overviews and AI Mode is included in the Search Console Performance report under the Web search type. Since 2026 there is also a dedicated generative AI performance report showing impressions, pages, countries, devices and dates; Google's documentation does not list clicks in that report. Its impressions are already counted in the main report, so they should not be added together.

Newsletter

New ideas, analysis and research — directly to your inbox.

Subscribe to receive new Insights publications and other selected content from Maccam Network. No spam. Unsubscribe at any time.

Shall we talk about your business?

Let's talk about what your business needs.

A 30-minute conversation is enough to understand the context, identify the problem and see if we are the right team to help you.