Every few months the marketing industry invents a new thing you are behind on. This year it is AI SEO. Also GEO. Also AEO. The acronym depends on who is emailing you, but the pitch is the same: Google works completely differently now, your SEO is obsolete, and there is a new service you need to buy immediately.
Google has published guidance on how AI Overviews and AI Mode pick the pages they cite. It is worth reading before you sign anything, because it is a lot less dramatic than the pitch.
AI SEO is not a separate technical ranking system. According to Google, websites do not need special AI files, special schema, or additional technical optimization to appear in AI Overviews or AI Mode. Pages still need strong content, crawlability, indexing, authority, and traditional SEO fundamentals.
What Is AI SEO?
AI SEO is the work of getting your pages used and cited in AI-generated answers, whether that is an AI Overview at the top of Google or a response in AI Mode.
The goal is real. Being the business an AI summary names is genuinely valuable. What is not real is the framing that this runs on a hidden second algorithm with its own file formats and its own consultants. GEO and AEO are other names for the same goal. They describe an outcome, not a separate technical discipline, and a lot of money is currently being charged for that distinction.
Does Google Require Special AI SEO?
No. In Google's official guide to AI features and websites, the company states plainly that there are no additional technical requirements to appear in these features.
Not "few." Not "minimal." None.
Google's position is that the same things making a page work in normal Search are what make it usable in an AI answer. Which is either disappointing or a relief, depending on whether you already paid for an AI SEO package.
How Does Google Choose AI Overview Sources?
Google lists what a page needs in order to show up as a supporting link:
- It has to be indexed by Google.
- It has to be eligible to appear in normal Search with a snippet.
- It has to meet Google's Search technical requirements.
Read the first two again, because they disqualify a lot of sites before the AI conversation even starts.
If Google has not indexed the page, nothing else matters. If near-identical pages got merged and Google indexed a different URL than you expected, that is an indexing problem, and fixes there can take up to two weeks to settle. If you suppress snippets, you have opted out of the mechanism these features draw from.
Most "we are invisible in AI" problems are ordinary technical SEO and content problems wearing a more fashionable hat.
What Is Query Fan-Out?
Google says these features may use a technique it calls query fan-out, issuing multiple related searches across subtopics and data sources to build a response. Google says this lets them surface a wider and more diverse set of links than a standard results page.
So one user question can become several searches behind the scenes. Somebody asks about hiring a contractor, and Google may go looking for pricing, timelines, licensing, and warranty terms as separate subtopics.
You are not fighting for one keyword slot anymore. You can get pulled in because you had the best page on one narrow piece of it. Good news for small businesses, and the opposite of what the enterprise AI tooling pitch implies.
Do You Need llms.txt or Special AI Schema?
Google's answer is no on both. The documentation says you do not need to create new machine-readable files, AI text files, or markup to appear in these features, and that there is no special schema.org structured data you need to add.
On llms.txt specifically: Google has not said it uses it. There is a difference between a file existing, some tools supporting it, and Google ranking you because of it. Anyone telling you an llms.txt file will make your site rank is describing a mechanism Google has not claimed.
We covered a version of this when the trend was building Markdown copies of every page for AI crawlers. Same shape, same problem: maintenance and duplicate-URL risk in exchange for a benefit nobody has demonstrated.
Structured data still matters. Google's guidance is that it be accurate and match what people can see on the page. That is a correctness requirement, not a secret AI unlock.
What Legitimate AI SEO Agencies Actually Do
There is real work here. It just is not mystical. Based on what Google says its systems draw from, honest work looks like:
- Finding the questions your customers actually search for, including the small ones.
- Building focused pages that answer those questions directly instead of burying the answer under company history.
- Making important content available as text. Google calls this out specifically, and an answer trapped in an image or untranscribed video is one Google cannot read.
- Fixing crawling, indexing, internal links, and page experience on real phones.
- Providing good images and video, accurate structured data, and a current Business Profile.
- Building brand authority, the slow part nobody wants to sell because it cannot be finished in ninety days.
- Publishing original examples, real pricing, and firsthand experience nobody can copy off a template.
Look at that list. It is a good SEO program. That is the point.
Can Anyone Guarantee Google AI Overview Rankings?
No, and four specific promises are worth treating as red flags:
- A specific AI Overview placement. Google decides whether an AI Overview appears at all, and says these show when they add value beyond standard results. That is not a slot anyone sells.
- That Google will cite a particular page. You can make a page a strong candidate. You cannot make it the pick.
- That an llms.txt file will make you rank. Google has not said this.
- Permanent inclusion. AI answers get regenerated. Being cited in June is not tenure.
Strong fundamentals make you eligible and competitive. They do not make you guaranteed, and no honest agency will say otherwise. The same caution applies to ordinary ranking movement after a core update.
Why a Focused Article Beats a Bigger Website
Picture a construction company with a perfectly nice website. Home, services, gallery, about, contact. It ranks for its own name. It generates almost nothing.
Now picture one article on that same site called "Why does my construction website rank but not get calls," which actually answers it: the site ranks for the company name rather than the job, the phone number is an image, the service area is never stated in text, and the quote form asks for eleven fields.
That article can get pulled into AI answers the homepage will never touch, because it covers one subtopic completely. The homepage is about a company. The article is about a problem somebody typed into Google at eleven at night.
A smaller site with five genuinely useful pages regularly beats a bigger site with fifty vague ones. That was true before AI Overviews. Fan-out raised the payoff.
How Small Businesses Can Improve AI Search Visibility
Concrete steps, none of which require a new subscription:
- Read your actual Search Console queries. Not what you assume people search. What they typed. Most owners have never looked.
- Search your key questions and inspect the AI Overview citations. See who got cited and what kind of page it was. It is usually a focused answer page, not a homepage.
- Identify the subquestions inside the answer. Those are the fan-out subtopics, and that list is your content plan, sourced from Google rather than a keyword tool.
- Build the better, more complete resource. Not longer. More complete, and easier to pull a clear answer out of.
- Add what only you have. Real jobs, real numbers, real photos, real limitations, real prices.
- Measure conversions, not mentions. Getting cited is nice. Getting called is the business. We dug into what AI Overviews did to real click-through rates, and visibility and revenue have drifted apart.
That last one matters most. You can get more visible in AI answers and not make more money, and without lead tracking you will never know which happened.
AI SEO Without the Buzzword Nonsense
Our position: AI search visibility is real and worth pursuing. AI SEO as a separate product with its own invoice mostly is not.
The work that earns AI visibility is the work that earns regular search visibility. Solid technical foundations, content that answers real questions, pages Google can crawl and index, accurate structured data, and genuine authority. Google's Search Central AI Overview documentation points at fundamentals, not a secret file you forgot to upload.
So when someone quotes you a monthly AI SEO retainer, ask what specifically they will do that is different from good SEO. If the answer is special files, special markup, or guaranteed placements, you now know more than the pitch does. Panicking about technical minutiae is its own trap, which we got into when Google said CMS-injected junk URLs generally do not hurt you.
Miami Web Lab treats this as one part of a whole. SEO, content, technical health, tracking, and conversion, wired together, so you can tell whether any of it produced customers. If you want to know where your site stands, we will look at it and tell you straight, including if AI visibility is not actually your problem.