How to Improve AI Search Visibility: What Google Actually Published
Google has now written down how its AI features choose sources, which GEO tactics to ignore, and how to measure the result. Here is what the documents say and how I apply them for small businesses.
Written by David Cragg, Tucson SEO consultant, in search since 1990. Published May 23, 2026. Last reviewed September 17, 2026 against Google's current documentation.
The short version
- There is no second algorithm to game. Google says there are no additional requirements and no special optimizations needed to appear in AI Overviews or AI Mode.
- Google now names the tactics to skip. llms.txt files, content chunking, rewriting pages for AI, chasing inauthentic mentions and overfocusing on schema are all on its list.
- One setting matters. Search Console has a Search generative AI control, live for every site since August 31, 2026. Include is the default. Check that nobody switched it off.
- Measurement exists now. The Search Console generative AI report shows impressions in AI Overviews and AI Mode. It does not show clicks.
- ChatGPT, Claude and Perplexity are invisible to Search Console. Those visits show up as referral traffic in GA4.
On this page: What it means · What Google published · The one setting · What to skip · Non-commodity content · How to measure it · Local businesses · Questions
What AI search visibility actually means
AI search visibility is how often your pages are pulled into an AI-generated answer and linked as a source, instead of appearing only as a blue link the searcher has to click.
It is a different outcome from a ranking position, but it comes from the same work. Google's AI answers are built on its ordinary search index and ranking systems. The pages that earn organic visibility are the pages that get quoted.
I have been in search since 1990. Every time the results page changes shape, the industry invents a new specialty with a new invoice attached. For about two years AI search had no primary source at all, so everybody guessed and charged for the guess. That changed in 2026, when Google published its own rules.
What Google published in 2026
In May 2026 Google added a guide called Optimizing your website for generative AI features on Google Search to the fundamentals section of Search Central, and announced it on the Search Central blog. It takes about fifteen minutes to read. It will save you from most of the sales calls you are getting.
The guide describes two mechanics in plain terms. The first is retrieval-augmented generation: the AI pulls current pages from the Search index and grounds its answer in them. The second is query fan-out: the model splits one question into several related searches and gathers results for each. Both run on Google's core ranking systems. That is why Google says ordinary SEO still applies.
The guide also takes a position on the new labels. Google treats "AEO" and "GEO" as SEO under another name, and points readers who are considering those services to its guidance on evaluating third-party SEO advice.
The companion page, AI features and your website, is blunter still. It says a page must be indexed and eligible to show with a snippet, and that no other technical requirements apply.
That sentence gets misread in both directions. It does not mean AI search is unimportant. It means there is no separate door. Anyone selling you a separate door is selling something Google says does not exist.
The one setting worth checking today
There is one new control, and it lives in Search Console, not in your site files. Under Settings, the Search generative AI control decides whether your content can appear in AI Overviews, AI Mode and the generative features in Discover. Google finished rolling it out to every website on August 31, 2026.
Include is the default, so most sites need to do nothing. But the setting is inherited from parent properties. If a developer, an agency or a past employee changed it on the domain property, every page underneath follows. Opening the page takes thirty seconds.
Google says the setting is not used as a ranking signal for the rest of Search. It is also separate from AI training, which Google handles through its Google-Extended token. Robots.txt and crawler decisions are covered in my guide to optimizing a site for AI-driven search.
What Google tells you to skip
This is the part most agencies have not caught up with. The guide has a section on things you can ignore, and several of them are products currently sold to small businesses.
| Tactic | How it gets sold | What Google's guide says |
|---|---|---|
| An llms.txt file | A menu of your best content for AI systems | Google Search does not use it. It neither helps nor hurts your visibility in Search. |
| Content chunking | Break every page into small blocks so an AI can lift them | No requirement. Google can find the relevant part of a page that covers several topics. |
| Rewriting pages for AI | Add every long-tail phrasing a person might type | Not needed. The systems understand synonyms and meaning. |
| A page for every fan-out query | Publish dozens of pages covering each question variation | Doing this to manipulate AI answers violates the scaled content abuse policy. |
| Paid or seeded mentions | Get your brand named on listicles so the models learn it | Inauthentic mentions are less helpful than they look, and spam systems filter what they can. |
| Schema as an AI lever | Special markup that makes the AI quote you | No special schema.org markup is needed. Keep using it for rich results. |
| Tools claiming inside data | A dashboard built on Google's internal AI metrics | No third-party tool has access to Google's internal ranking or AI systems. |
On llms.txt, the outside data agrees with Google. An Ahrefs study of 137,000 domains found that about 28% published the file, and 97% of those files received no requests at all in May 2026.
Look at the pattern in that table. Everything on the skip list is cheap, fast and easy to invoice. What Google tells you to do instead is slow, specific and hard to automate. I have watched the same pattern play out with every SEO fashion since the early web directories.
Non-commodity content is the actual instruction
Google says unique, useful content will probably shape your presence in AI search more than anything else in the guide. It calls this non-commodity content: material based on first-hand experience or real expertise, rather than a summary of what is already online.
Google's own example contrasts a generic list of tips for first-time homebuyers with a first-hand account of waiving an inspection and what turned up in the sewer line. Translated to a local trade, the difference looks like this:
- Commodity: "7 signs you need a new AC unit." Anyone with a search engine could write it, and an AI can write it without you.
- Non-commodity: what a technician actually finds inside rooftop units after a Tucson monsoon season, with photos, the repair that was done and what it cost.
This is an editorial instruction, not a technical one. It is also the main reason a one-person shop can still beat a content farm. The details only the owner knows are the details a model cannot generate. Getting those details out of a busy owner and onto the page is most of what my content work is for.
How to measure AI search visibility
Until mid-2026 there was no honest answer to "am I showing up in AI results?" People treated screenshots of ChatGPT as data. There are now three checks, and I use all three.
1. The Search Console generative AI report
Google introduced the generative AI performance reports in June 2026, and the Search version reached all websites on August 31, 2026. It counts an impression each time a link to your site appears in AI Overviews or AI Mode. Discover has a separate generative AI report.
You can break impressions down by page, country, device and date. The help page does not describe click or query data. If the report is empty, your site may not have enough AI impressions yet, or the generative AI control may be set to exclude.
2. A GA4 referral segment
Search Console only sees Google. Someone who asks ChatGPT, Claude, Perplexity or Copilot for a recommendation and then clicks through arrives as ordinary referral traffic. In GA4, build an exploration filtered on session source with a pattern such as chatgpt|perplexity|claude|copilot|gemini, and check it monthly.
On small local sites the volume is usually tiny next to organic search. Judge it on conversions instead. A visitor who was just told you are the answer tends to arrive ready to call.
3. Prompt checks, recorded as anecdotes
Once a month I ask the assistants what a customer would ask, in the customer's words. Not "best HVAC Tucson," but "my AC quit and it's 108 degrees, who can come today in Tucson."
I write the results down as anecdotes, never as rankings. The same prompt asked twice can name different businesses. One screenshot of a competitor proves very little, and one screenshot of you proves just as little.
What this means for a local business
Local searches are where AI answers hit hardest, because a generated answer often names two or three businesses and stops. There is no page two to scroll to.
Google's guide says Business Profile information can help your services appear in AI responses as well as regular results. A complete, accurate profile is doing double duty now, and keeping it that way is the job of Google Business Profile management. How the three Map Pack spots get chosen is covered on my Tucson SEO services page.
One point is specific to AI answers: the words customers use in reviews become part of how these systems describe you. A review that says "replaced our condenser in Marana the same afternoon" tells an answer engine more than five reviews that say "great service." You should never script reviews. You can ask better questions when you request one, and reputation management is where I set that up.
The order I work in
- Confirm the page can be found. Indexed, eligible for a snippet, and the Search generative AI control set to include. If a crawler cannot reach the page, nothing else matters. The site-side checks are in the four-gate readiness guide.
- Answer the question in the first sentence. Not after a paragraph of setup. People skip that setup, and so do answer engines.
- Give each question its own page. A page that tries to serve two intents answers neither clearly. Reading search intent is where this starts.
- Add what only you could write. A real job, a real number, a real thing that went wrong.
- Name the author, and mean it. A page signed "the team" carries no experience signal.
- Measure monthly. Check the Search Console report, the GA4 segment and a short prompt log.
Notice what is missing: no new file in the site root, no reformatting every paragraph into fragments, no buying mentions. How this plays out differently by trade is covered in industry-specific SEO.
Who wrote this
I'm David Cragg. I have worked in search since 1990, hold a BA in Economics from UCLA and an MBA from the University of Washington, and founded and sold two internet marketing companies, Lotus411 and MSD2D. Today I work with small businesses across Tucson and Southern Arizona: one consultant, $1,000 a month, no contract.
I use AI tools to draft and to check my work. They do not supply facts, and nothing goes up here that I have not read and checked against the sources below. More on my background, how the engagement works and results for other businesses.
Questions about AI search visibility
What is AI search visibility?
AI search visibility is how often your pages are pulled into an AI-generated answer and linked as a source, instead of appearing only as a blue link. Google builds its AI Overviews and AI Mode answers on its ordinary search index and ranking systems, so the pages that earn organic visibility are the pages that get quoted. Treat it as a second payoff from solid SEO, not a separate product you buy on the side.
Do I need to do anything special to appear in Google's AI Overviews or AI Mode?
No. Google's documentation says there are no additional requirements to appear in AI Overviews or AI Mode, and no other special optimizations are needed. A page must be indexed and eligible to show with a snippet, and your site must be included under the Search generative AI control in Search Console, which is the default. Google also says no special schema.org markup is required. Structured data is still worth using for rich results, as long as it matches the visible text.
Is an llms.txt file worth adding to my site?
Not for Google. Google's generative AI guide says Google Search does not use llms.txt files, and that publishing one will neither help nor hurt your visibility in Search. Outside data points the same way: an Ahrefs study of 137,000 domains found that 97% of published llms.txt files received no requests at all in May 2026. If another tool you use reads the file, publishing it is harmless. It is not an AI search strategy.
How do I measure whether my pages appear in AI answers?
Use three checks. The Search Console generative AI performance report, available to all sites since August 31, 2026, counts impressions when links to your site appear in AI Overviews or AI Mode, broken down by page, country, device and date. A GA4 referral segment catches visits from ChatGPT, Claude, Perplexity and other assistants. A monthly log of the prompts your customers would use shows the competitive picture, as long as you treat it as anecdote rather than a ranking.
Does Google Search Console show traffic from ChatGPT, Claude or Perplexity?
No. Search Console only measures Google Search and Discover. Visits that start in ChatGPT, Claude, Perplexity, Copilot or the Gemini app arrive as ordinary referral traffic and appear in Google Analytics 4 under the referring domain. Neither tool gives the whole picture, so anyone quoting you a single AI visibility percentage is estimating.
What does Google mean by non-commodity content?
Non-commodity content offers first-hand experience or expert insight that goes beyond common knowledge. Commodity content restates what is already online and could be written by anyone, or by an AI model. A page called "7 signs you need a new AC unit" is commodity content. A page showing what a technician found inside rooftop units after a Tucson monsoon, with photos and the repair cost, is not. Google says this kind of content likely matters more for AI search than anything else in its guide.
Can I keep my site out of AI Overviews and AI Mode?
Yes. The Search generative AI control in Search Console, live for all websites since August 31, 2026, lets you exclude your site's links and content from AI Overviews, AI Mode and generative features in Discover. Excluded sites get no traffic or impressions from those features. Google says the setting is not used as a ranking signal elsewhere in Search and does not control AI training. For most local businesses excluding makes little sense, because those answers are where customers now ask for recommendations.
How often do AI answers change?
Far more often than blue-link rankings. These systems retrieve fresh sources and generate a new answer each time, so the same prompt asked twice in an afternoon can name different businesses. Track Search Console impressions over a month instead of reacting to a single result, and be skeptical of any tool that reports an AI ranking as one fixed number.
Should I pay for an AEO or GEO service?
Be careful. Google says optimizing for its generative AI features is still SEO, and it points anyone considering AEO or GEO services to its guidance on evaluating third-party SEO advice. Ask a vendor which of their tactics appear on Google's skip list, such as llms.txt files, content chunking or seeded mentions, and whether they claim access to Google data no outside tool has. If the package is mostly those items, you are paying for work Google says to skip.
Want to know if you're showing up?
I will check your Search Console generative AI report, your GA4 referrals and your main service pages, and tell you plainly whether the problem is AI visibility or something more ordinary. Usually it's something more ordinary.
Primary sources
Everything above can be checked. These are the documents this page relies on:
- Optimizing your website for generative AI features on Google Search, Google Search Central, May 2026
- AI features and your website, Google Search Central
- Search generative AI control, Search Console Help
- Generative AI performance report (Search), Search Console Help
- Introducing Search generative AI performance reports in Search Console, June 2026
- Ahrefs llms.txt study of 137,000 domains, June 2026
AI search changes faster than any page about it. If a detail here has aged, call me and I will tell you what changed.