Google’s 2026 AI SEO Guidance: What You Don’t Actually Need to Do

I’ve spent a fair amount of time over the last year trying to make sense of what “AI SEO”, GEO, AEO and all the other variations of the acronym soup actually mean for a business website.

And I’ll admit: some of the advice out there can make it sound as though we’re supposed to rebuild the internet every six months.

Create special files.

Rewrite everything for AI.

Break articles into tiny chunks.

Add special AI schema.

Get your brand mentioned everywhere.

Buy an “AI visibility” score.

The list goes on.

So when Google published its new “Optimizing your website for generative AI features on Google Search“ guidance in May 2026, I paid particular attention to one section.

It’s called:

“Mythbusting generative AI search: what you don’t need to do.”

That’s a pretty useful heading.

The guidance was published by Google Search on 15 May 2026 and was subsequently updated. Google says the purpose is to help website owners, SEOs and developers understand how to optimise for generative AI features in Google Search and, importantly, what they can ignore.

And there are some things in there that I think businesses paying for SEO or “GEO” services should know about.

Because if somebody is charging you extra for something Google explicitly says isn’t necessary for Google Search, it’s reasonable to ask exactly what you’re paying for.

First, the big takeaway: Google says SEO still matters

Before getting into the mythbusting, there’s an important point that could easily get lost.

Google isn’t saying:

“Forget SEO. AI has changed everything.”

Quite the opposite.

Google’s current guidance says that its generative AI features, such as AI Overviews and AI Mode, are rooted in its existing Search ranking and quality systems.

Google describes these systems as using techniques including retrieval-augmented generation (RAG) and query fan-out to find relevant information from the Search index and use it to construct responses.

In other words, for Google Search at least, the advice is remarkably familiar:

Make your website crawlable.
Get the important pages indexed.
Create useful content.
Make the content genuinely valuable.
Make the website usable.
Keep your business information accurate.

There isn’t a secret second version of SEO that you have to unlock because an AI is involved.

And that’s probably the most useful thing in the entire document.

So what doesn’t Google think you need?

Google specifically addresses several of the tactics that have become popular under the AEO/GEO banner.

Let’s look at them.

1. You don’t need an llms.txt file for Google Search

This is probably the one I would put at the top of the list.

There has been plenty of discussion around llms.txt — a proposed file intended to give AI systems a simplified representation of a website.

And there are now tools and services that will generate these files for you.

But Google’s position is very clear.

You do not need to create llms.txt, special AI text files, machine-readable files, Markdown files or other special markup to appear in Google Search or its generative AI features.

Google says Search doesn’t use them for this purpose. It also says that maintaining one for another service or system is fine, but it won’t help or hurt your visibility or rankings in Google Search.

That’s an important distinction.

If someone is selling you an llms.txt file because it is supposed to make your website more visible in Google AI Overviews or AI Mode, Google’s own documentation doesn’t support that claim.

If they’re recommending it for another AI system, documentation platform, AI agent or a particular technical workflow, that’s a different conversation.

But:

“You need llms.txt for Google AI search” is not supported by Google’s current guidance.

Google even clarified its llms.txt position in June 2026, stating that these files aren’t needed for Google Search and won’t positively or negatively affect visibility or rankings there.

That’s quite a useful thing to know before paying a monthly subscription for one.

2. You don’t need to “chunk” your content for AI

Another popular idea has been that websites need to break their content into small, highly isolated chunks so that AI systems can retrieve individual passages more effectively.

Google doesn’t say you need to do this.

Its guidance specifically says there’s no requirement to break content into tiny pieces for AI to understand it.

Google says its systems can understand nuance across multiple topics on a page and retrieve the relevant part when appropriate.

And there is no magic page length.

Short pages can work.

Long pages can work.

The appropriate length depends on the subject and the audience.

This doesn’t mean structure doesn’t matter.

Of course it does.

Good headings, sensible sections, descriptive links and readable content are all good web design and SEO practice.

But there’s a difference between:

“Make your content easy for humans to navigate.”

and:

“Break every article into 300-word AI retrieval chunks because Google requires it.”

The first is sensible.

The second isn’t supported by Google’s current guidance.

3. You don’t need to rewrite your content specifically for AI

This one is particularly interesting. I have even dedicated a whole page on my site to this.

That’s because there’s been a lot of advice suggesting that websites need to rewrite their content in a special format for generative AI.

More conversational sentences.

More question headings.

More long-tail variations.

More “answer-first” formatting.

More repetitions of the same entity and keyword.

Some of these techniques can be useful from a human usability perspective.

But Google says you don’t need to write in a special way specifically for generative AI Search.

Its systems can now understand synonyms and general meanings rather than requiring an exact keyword match for every possible variation of a search query.

That’s why we’re changing tune again (every six months, as I’ve mentioned…).

Now you don’t need to write:

“Best website designer in George”

into every conceivable variation:

“Best website designer George”

“George best website designer”

“Website design best George”

“Website designer near George”

until the page sounds like it was written by a malfunctioning keyword generator.

Write for the person.

Explain the subject properly.

Use terminology that makes sense.

Cover the relevant questions.

Don’t obsess over manufacturing every possible long-tail keyword variation.

4. You don’t need special “AI schema”

This one needs a little more nuance.

Google says structured data is not required for generative AI search, and there is no special Schema.org markup that you need to add specifically to make your website appear in AI Overviews or AI Mode.

That does not mean schema is useless.

Far from it.

Structured data can still help Google understand your content and can make pages eligible for supported rich results and other search features. Google’s general documentation continues to describe structured data as a way of helping Search understand page content and potentially enabling richer search appearances.

So I’d make the distinction:

Don’t pay for “AI schema” because you think Google requires it for AI search.

But don’t go through your website deleting useful, correctly implemented structured data either.

For a WooCommerce site, for example, product structured data can still be relevant.

For a local business, appropriate business information can still be useful.

For other supported search features, structured data can still matter.

The point is simply that there isn’t a secret AI Schema™ checkbox hiding somewhere in WordPress.

5. Don’t manufacture “mentions” of your business

This one requires more careful wording.

Google acknowledges that its generative AI features can use information about businesses, products and services from across the web—including blogs, videos and forum discussions.

So yes, being genuinely discussed elsewhere can be useful.

But Google specifically warns against trying to manufacture inauthentic mentions simply because you believe AI systems will notice them. Its ranking systems focus on high-quality content, while its spam systems are designed to deal with spam.

There’s a fairly important difference between:

“Let’s get our business genuinely mentioned by relevant organisations, customers and publications.”

and:

“Let’s manufacture 300 mentions of this company across random websites because an AI visibility tool says we need them.”

The first is normal marketing.

The second starts looking like an attempt to manipulate the system.

And Google explicitly says its spam policies apply to generative AI responses in Search too.

So I’m not throwing out digital PR, genuine reviews, partnerships, citations or legitimate third-party coverage.

Quite the opposite.

I’m saying:

Earn the mentions. Don’t manufacture them.

6. Don’t believe anyone who says they’re “Google approved”

This is perhaps the most useful section for anyone considering paying for SEO or GEO services.

Google has now published separate guidance specifically about evaluating third-party SEO tools, services and advice.

And it makes an important point:

Google doesn’t evaluate or endorse third-party SEO services or tools.

So if somebody tells you that their GEO platform, AI visibility software or SEO system is “Google approved”, that’s something I’d want to see documented very carefully.

There’s an even bigger point.

Google says:

Third-party tools don’t have access to Google’s internal ranking data.

And:

They can’t guarantee performance.

Google also says that predictions made by these tools are their own predictions and may not happen.

That doesn’t mean SEO tools are useless.

I use tools.

Most experienced SEO professionals use tools.

They can be extremely useful for research, crawling websites, identifying technical issues, analysing competitors, monitoring keywords and finding opportunities.

But you need to understand what the numbers actually represent.

Your “AI Visibility Score” isn’t a Google score

This is one of the things I’d pay particular attention to.

If a platform gives you an:

AI Visibility Score: 73/100

that number may be useful as that platform’s own measurement.

But it isn’t a measurement coming from Google’s internal AI ranking systems.

It’s a model.

The same applies to various visibility, authority, sentiment, citation or “share of AI” metrics produced by third-party platforms.

They can be useful for comparison and trend monitoring.

But don’t confuse:

“This tool estimates your visibility at 73.”

with:

“Google says your AI visibility is 73.”

Those are completely different statements.

And Google has now explicitly told website owners to be cautious about third-party tools making claims about internal Google metrics or guaranteed performance.

Interestingly, Google also now provides a first-party way of measuring visibility in its own generative AI features through Search Console’s Generative AI performance report.

That’s a much more useful starting point when we’re specifically talking about Google’s own AI Search experiences.

What Google actually wants us to focus on

And here’s the slightly amusing part.

After all the talk about GEO, AEO, AI visibility and optimisation for generative search, Google’s actual recommendations are remarkably… boring.

And I mean that as a compliment.

Google says to focus on:

Useful, non-commodity content

Google places particular emphasis on content that provides something genuinely useful and distinctive.

That could be:

First-hand experience
Original research
Case studies
Unique data
Expert insight
A genuinely different perspective
Detailed examples
Useful local knowledge
Original images or video

Google specifically contrasts unique, experienced content with generic material that simply repeats information already available elsewhere.

Make your content easy to use

Use headings.

Use paragraphs.

Structure information sensibly.

Make pages readable.

Answer the user’s question.

Don’t make people dig through 2,500 words to find the one sentence they actually needed.

Notice that this doesn’t require some mysterious AI-writing formula.

It’s simply good web publishing.

Keep the technical foundations healthy

Google’s AI features still depend on the Search index.

That means the fundamentals still matter:

Crawling
Indexation
Internal linking
Search technical requirements
Mobile usability
Page experience
Accessible content
JavaScript SEO where relevant
Avoiding unnecessary duplication

Google specifically says that a page needs to be indexed and eligible to appear in normal Google Search in order to be eligible as a supporting link in AI Overviews or AI Mode. There aren’t additional technical requirements just for those AI features.

That’s a pretty important point.

If Google can’t properly access your website in the first place, an expensive GEO strategy isn’t going to rescue it.

Local SEO hasn’t suddenly disappeared either

For local businesses, Google’s guidance is particularly relevant.

Google specifically calls out Google Business Profile as part of maintaining local business visibility.

It also mentions Merchant Center and product information for relevant businesses.

So for a local business, the boring checklist remains remarkably important:

Is your Google Business Profile accurate?

Is your address correct?

Are your opening hours correct?

Are your services properly represented?

Does your website agree with your business information?

Are customers leaving genuine reviews?

Are your important local pages useful?

Is your website crawlable and indexable?

That’s much more actionable than worrying about whether you’ve created a special AI text file.

So what changed since my January 2026 SEO articles?

This is where things get particularly interesting.

I wrote two articles earlier in 2026 looking at the changing SEO landscape:

SEO for 2026
How to Survive the 2026 SEO Shift

At the time, I was looking closely at the ideas around information gain, content redundancy and what I referred to as Google’s emerging “GIST” approach.

Some of the underlying thinking still makes sense.

But after reading Google’s current documentation, I’d refine some of my earlier conclusions.

And that’s exactly what SEO should involve.

You change your understanding when better evidence becomes available… again, the six-months reference.

What still holds up

The broad direction of my January thinking remains useful.

I still believe businesses should move away from:

Generic SEO articles
Keyword stuffing
Rewriting whatever already ranks
Mass-produced AI content
Thin location pages
Content created purely to generate traffic
Anonymous, generic expertise claims

Google’s current guidance strongly supports the underlying principle of creating unique, useful, non-commodity content rather than simply recycling information that already exists.

So the principle of “give people something worth finding” remains.

What has changed is the degree of certainty we should attach to some of the mechanisms I discussed earlier.

What I would change from my January articles

One of my earlier articles described GIST as though it were an established Google algorithm and discussed concepts such as a “Vector Exclusion Zone” and content being filtered because it provided no additional information.

Having now gone through Google’s current published guidance, I would be more careful with that language.

Google’s current public documentation does not establish GIST as a named Google Search algorithm with those mechanisms.

Google does explicitly discuss the importance of unique, non-commodity content and avoiding content created primarily to manipulate search or generative AI responses.

That’s a meaningful distinction.

So I would now describe the strategy as:

Create genuinely useful, distinctive content that adds something to the conversation.

rather than claiming:

Google’s GIST algorithm will put your content into a specific exclusion zone if it doesn’t.

The first is supported by Google’s published guidance.

The second goes beyond what Google has publicly documented.

That’s an important correction.

What about the “answer-first” approach?

I would keep it but for a different reason.

In my earlier work, I suggested that giving Google a concise summary or “gist” of a page could make it more suitable for AI-generated answers.

There’s nothing wrong with doing that.

In fact, a clear introduction is often excellent UX.

But I wouldn’t tell a client:

“You need a 100–150 word answer at the top because Google’s AI requires it.”

Google’s guidance doesn’t say that.

Instead:

Give the reader the useful information quickly because that’s good content design.

If that also makes the content easier for search systems to understand and retrieve, excellent.

The motivation should be the user—not a made-up formatting requirement.

And what about schema?

This is another area where I’d refine my previous advice.

My January SEO material placed fairly strong emphasis on schema as part of the modern search strategy.

I still recommend appropriate structured data where it makes sense.

But Google’s new guidance makes the distinction much clearer:

Schema isn’t required specifically for generative AI Search.

It remains useful for supported Search features and rich results.

So the strategy isn’t:

“Add as much schema as possible for AI.”

It’s:

“Use accurate structured data where it helps Google understand the page or makes the site eligible for supported search features.”

That’s a much healthier approach.

What should businesses focus on right now?

If I were reviewing a client’s website today, this is where I’d put my attention.

1. Get the technical basics right

Before worrying about GEO:

Is the site crawlable?
Are important pages indexed?
Are there accidental noindex directives?
Is robots.txt blocking anything important?
Are canonical URLs correct?
Are internal links sensible?
Are pages mobile-friendly?
Is the site reasonably fast?
Are Core Web Vitals and overall page experience being monitored?
Are important assets accessible to search engines?

Google’s current guidance makes it very clear that technical accessibility remains foundational to AI Search visibility.

2. Improve the content you already have

Don’t automatically start producing 50 new blog posts.

Look at what you’ve already published.

Ask:

Is this actually useful?

Is it accurate?

Does it demonstrate experience?

Does it answer the question properly?

Is there anything here that couldn’t have been generated from five minutes of generic research?

Could I add a real example from my business or customers?

Sometimes the best SEO project isn’t creating another article.

It’s making the existing one substantially better.

3. Add experience

This is one of the areas I think small businesses have an enormous opportunity.

Show your work.

Explain what you’ve learned.

Publish case studies.

Use original photography.

Share customer experiences.

Document projects.

Give examples.

Talk about mistakes you’ve made and what you learned from them.

Explain how you actually solve problems.

That material is difficult to manufacture because it comes from having actually done the work.

And Google explicitly encourages unique perspectives and first-hand experience as examples of non-commodity content.

4. Don’t manufacture authority

Don’t buy 500 meaningless mentions because an AI tool says you need them.

Don’t create fake reviews.

Don’t publish generic guest posts purely to insert brand mentions.

Don’t create dozens of pages targeting every possible variation of a query.

Build genuine visibility instead.

Get involved in your industry.

Work with real partners.

Earn genuine reviews.

Publish useful material.

Get talked about because you’ve actually done something worth talking about.

5. Keep your local presence accurate

If you’re a local business, maintain your Google Business Profile.

Keep your business information consistent.

Make sure your services and products are represented accurately.

Collect genuine customer reviews.

Create useful local content where appropriate.

Google specifically identifies Business Profile information as relevant to local visibility in its generative AI guidance.

6. Use Search Console as your source of truth

Third-party tools have their place.

But when we’re talking about Google Search, I want to see what Google itself is telling us.

Search Console can show:

Search queries
Impressions
Clicks
CTR
Search positions
Indexing information
Technical problems
And now, generative AI performance data

Google specifically recommends its first-party Search Console data when evaluating performance, while warning that third-party tools don’t have access to its internal ranking data.

That doesn’t make third-party tools bad.

It just means we need to understand what they’re measuring.

One important caveat: Google isn’t ChatGPT

This is probably the most important qualification to Google’s guidance.

Everything I’ve discussed above is about Google Search’s own generative AI experiences, particularly AI Overviews and AI Mode.

It is not a universal statement about every AI system on the internet.

ChatGPT, Claude, Perplexity and other AI products can use different models, retrieval systems, indexes, tools and sources.

So if someone recommends an llms.txt file specifically for another technical ecosystem, AI agent, documentation system or application, that recommendation needs to be evaluated on its own merits.

Google is telling us what Google Search needs.

It’s not telling us how every AI system works.

That’s why I wouldn’t turn this into another simplistic: “GEO is dead.” headline. That’s not what Google said.

What Google did say is much more useful:

For Google’s Search ecosystem, you don’t need a special collection of AI optimisation tricks layered on top of SEO.

Until Google changes the guidance again, here’s my 2026 SEO checklist

If I had to reduce all of this to a practical checklist for a business owner, it would look something like this:

Do this

✓ Create useful, original, non-commodity content

✓ Demonstrate genuine experience

✓ Keep important content crawlable and indexable

✓ Maintain strong technical SEO

✓ Make pages useful and easy to navigate

✓ Use appropriate structured data

✓ Keep Google Business Profile information accurate

✓ Keep Merchant Center information accurate if you’re an ecommerce business

✓ Use original images and video where they genuinely add value

✓ Monitor performance in Google Search Console

✓ Improve existing content rather than endlessly producing more

✓ Build genuine third-party recognition

Don’t do this just for Google AI Search

✗ Don’t create llms.txt because someone says Google requires it

✗ Don’t create special AI files

✗ Don’t chop every article into tiny chunks

✗ Don’t rewrite perfectly good content purely to make it “AI readable”

✗ Don’t buy manufactured mentions

✗ Don’t pay for “special AI schema” because Google requires it

✗ Don’t treat an AI visibility score as an internal Google metric

✗ Don’t believe anyone claiming their SEO/GEO service is Google-approved without very strong evidence

The funny thing about Google’s AI SEO advice

After all the noise around GEO, AEO and AI optimisation, Google’s current advice is almost aggressively unexciting.

Make your website accessible.

Make your content useful.

Give people something they can’t get everywhere else.

Show your experience.

Keep your business information accurate.

Don’t spam.

Measure what you can actually measure.

That’s it.

And perhaps that’s the biggest lesson here.

The SEO industry has a natural tendency to turn every change in search into a new product category.

Sometimes that’s justified.

Sometimes there really are new technical requirements.

But Google’s May 2026 guidance is effectively telling website owners:

Don’t confuse a new search interface with an entirely new set of ranking rules.

For Google Search, the foundations still matter.

And until Google publishes something that changes that guidance, that’s where I’d be spending my client’s money.

Not on an AI-shaped layer of SEO theatre.

On the website.