If your customers use Google, they are already reading AI-generated answers before they ever reach a list of blue links. AI Overviews now sit above the results for a large share of queries, and AI Mode turns search into a back-and-forth conversation. For most businesses this raises one urgent question: how do we get our website into those answers, and win the click that follows?
There is a lot of noise about this. A whole vocabulary has sprung up, from "AEO" to "GEO" to "answer engine optimisation", along with a shopping list of supposed hacks. Much of it is wrong, and some of it is a waste of budget. This guide cuts through it using Google's own published guidance from 2026, and our experience applying it on live client sites.
The short answer
To appear in Google's AI Overviews and AI Mode, you need the same things that make you rank well in ordinary Google Search, because the AI features are built on top of Google's core Search systems. In practice that means four things: content that is genuinely useful and based on first-hand experience rather than recycled from elsewhere; a site that is technically crawlable, indexed and eligible to show a snippet; clear answers that are easy to extract; and real authority and trust in your subject. There is no separate "AI algorithm" to game, and most of the "GEO hacks" circulating online do nothing for Google.
The rest of this article explains why, and what to actually do.
What AI Overviews and AI Mode really are
It helps to be precise, because the two features behave differently.
AI Overviews are the AI-generated summaries that appear at the top of a normal results page for many queries. They answer the question directly and include clickable links to the sources the answer draws on.
AI Mode is a more conversational, end-to-end generative experience. Instead of a single answer, you can ask follow-up questions and go deeper, with the results assembled dynamically from across the web.
Two mechanisms sit underneath both, and understanding them changes how you write.
The first is grounding, also called retrieval-augmented generation. Rather than answering from memory, Google's AI features retrieve relevant, up-to-date pages from the Search index and build the response from what those pages actually say, then link to them. Your page can only be used if it is in the index and retrievable in the first place.
The second is query fan-out. For a single question, the system quietly issues a set of related sub-queries and pulls results for each. Ask "how do I fix a lawn full of weeds," and behind the scenes it may also look up weed killers, chemical-free removal and prevention. The practical upshot is that one strong, comprehensive page can be surfaced across a whole cluster of related sub-questions. You do not need a separate thin page for each variation, and as you will see below, creating them can actively work against you.
The truth nobody selling "GEO" wants to say: this is still SEO
Google's position on this is unusually blunt. Its 2026 guidance states that, from Search's point of view, optimising for generative AI features is optimising for Search. It is still SEO. The AI features are rooted in the same ranking and quality systems that decide ordinary results. There is no separate door to walk through.
That matters because it tells you where to spend. The fundamentals you may already be investing in, such as useful content, sound technical SEO and genuine authority, are exactly what drives AI visibility. If your foundations are weak, no amount of "AI optimisation" will paper over them. If you want to sense-check those foundations first, start with a proper SEO audit.
What actually works
1. Be genuinely non-commodity (this is the biggest lever)
Google is explicit that creating unique, useful content will influence your presence in generative AI search over the long run more than anything else on its list. The distinction it draws is between commodity and non-commodity content.
Commodity content is the stuff anyone could write from general knowledge: "7 tips for a faster website," "what is a CMS," a tidy summary of things already said on a hundred other sites. A generative model can produce that in seconds, which is precisely why it earns you nothing. There is no reason to retrieve your version.
Non-commodity content carries something only you have: a first-hand result, a number from a real project, a decision you made and what happened next, an opinion earned by doing the work. That is the material AI answers reach for when they need something specific and trustworthy, and it is the material competitors cannot clone.
The test we apply to every page we publish is simple. Would this still deserve to exist if Google got twice as good at spotting generic content? If the honest answer is no, we do not publish it. We explored this tension in more depth in AI content vs human content.
2. Make sure you're technically eligible
You cannot be cited in an answer you are not eligible to appear in. For a page to be usable by Google's AI features it must be indexed and eligible to be shown in Search with a snippet, so a nosnippet directive, for example, takes you out of contention. It must be crawlable, because the AI features rely on publicly accessible content, and blocked resources or heavy client-side rendering that Google cannot process will cost you. And the site must be included in Search's generative AI features, which is governed at the site level and visible in Search Console.
None of this is exotic. It is ordinary technical hygiene. But it is where we most often find the actual problem. It is worth confirming your Core Web Vitals and page experience are healthy while you are in there, and, in a mistake we see constantly, making sure staging sites and duplicate URLs are not quietly competing with your real pages.
3. Answer the question clearly, near the top
Because AI features extract answers, structure earns its keep. Lead with a clear, direct response to the core question, then expand. Use descriptive headings that match how people actually ask things. Break content into logical sections that a reader, and a retrieval system, can navigate.
Note what this is not. It is not "chunking" your content into tiny fragments for the machine (see the myths below). Google can understand multiple topics on a single page perfectly well. Write for a person who wants a clear answer, and the extractability follows.
4. Build real authority and trust
Google leans on a mix of signals associated with experience, expertise, authoritativeness and trust, with trust the most important of the four. It weights this even more heavily for "Your Money or Your Life" topics such as finance, health and safety, meaning anything that materially affects someone's life. If you operate in one of those areas, credibility is not optional.
Concretely, put real, named authors on your content with genuine bios. Make it obvious who is behind the site. Earn authentic recognition in your field rather than chasing it artificially. Where you have used AI to help produce something, be transparent about it wherever a reader would reasonably expect to know.
5. Support the words with images, video and structured data, in that order of effort
Google's AI features can surface images and video, so relevant, high-quality visuals give you more surface area to appear. Structured data is worth keeping for rich results in ordinary Search, but be clear-eyed about it. It is not required for AI features, and there is no special schema that unlocks them. Do it as part of good SEO, not as an "AI" tactic.
What to ignore (the myths Google has explicitly debunked)
This is where a lot of money is being wasted in 2026. Google has openly named several popular "GEO/AEO hacks" that do nothing for its Search features.
llms.txtfiles. Google Search ignores them. Keeping one will not help or hurt, so just do not expect visibility from it.- "Chunking" your content into tiny pieces for the AI to digest. This is unnecessary, because Google understands full pages.
- Rewriting content specifically for AI systems. You do not need a special robotic style or every long-tail phrasing variant. Google understands synonyms and intent.
- Chasing inauthentic "mentions" across the web to seem popular. Manufactured buzz is not the shortcut it is sold as.
- Treating structured data as an AI unlock. As above, it is good for rich results, but not a requirement for AI answers.
Our blunt take, from doing this work, is that if a tactic is framed as a clever loophole in how AI search "reads" your site, it is almost certainly a distraction from the two things that actually move the needle: better content and a sounder site.
How to measure whether it's working
Use the Generative AI performance report in Google Search Console to see how your content is doing in Google's AI features. It is the only source with a genuine line of sight into your own performance.
Be sceptical of third-party tools that promise "AI ranking scores" or claim access to Google's internal metrics. None of them have that access. Some are useful for workflow, so treat their numbers as estimates rather than truth, and check any advice against Google's official guidance.
Don't ignore the click problem
Here is the uncomfortable part. Ranking inside an AI answer is not the same as getting a visitor. For some questions, Google can answer completely and the user never leaves, which is the classic zero-click outcome.
That should change what you write about, not just how you write it. "What does SEO stand for" is a terrible commercial bet, because it is fully answerable in the result, so appearing there earns you nothing. "How much should a website redesign cost in the UK" is a strong one. An AI answer can give a range, but a serious buyer still wants your benchmarks, your examples, a calculator and a real case study, all of which are reasons to click through and, eventually, to enquire. Our SEO cost guide is a good example of a page built to earn the click after the answer, not just the impression.
Prioritise topics where there is a genuine reason to visit the source after the initial answer. Everything else is vanity visibility.
What this looks like in practice
The agencies and brands winning in AI search are not running a separate "GEO programme." They are doing excellent SEO on content that carries first-hand evidence. We see it in our own results, not just our clients'.
Take our guide to SEO pricing. It has climbed to the top of page one for its core terms, and Google now pulls several of its individual sections into AI Overviews, so the page is surfaced across a whole cluster of related "how much does SEO cost" questions. That is query fan-out working exactly as described above. A separate, deeply technical comparison of open-source headless CMS platforms tells the same story from the other end. Built on genuine hands-on detail rather than a generic overview, it reached page two on merit and earns a steady stream of clicks from a niche, high-intent developer audience. Neither page chases tricks. Both carry specifics a generative model could not invent, and that is precisely why they get pulled into answers.
The pattern, across everything we publish and every migration and rebuild we run, is consistent. The pages that earn a place in AI answers are the ones carrying first-hand detail, not the ones summarising what is already online. In our experience a genuinely well-evidenced resource tends to start appearing in AI features within weeks rather than months, but only when it says something the rest of the web does not.
"The teams asking us how to optimise for AI are usually asking the wrong question. There is no separate lever. We have never won an AI Overview placement with a clever trick. We win them by publishing something that was actually worth citing."
Rubik Digital
Your 2026 checklist
- Fix the foundations first, so you are crawlable, indexed, snippet-eligible and fast, with no duplicate or staging URLs competing with you.
- Confirm your site is eligible for AI features in Search Console, and watch the Generative AI performance report.
- Publish non-commodity content built on first-hand experience, results and data, not summaries of what is already online.
- Lead each page with a clear, direct answer, then structure it with descriptive headings.
- Put named experts and real trust signals on your content, especially for YMYL topics.
- Add relevant images and video, and keep structured data for rich results.
- Prioritise topics with a real reason to click through to you.
- Ignore
llms.txt, "chunking", AI-only rewriting, manufactured mentions and schema-as-AI-hack.
The bottom line
Visibility in Google's AI Overviews and AI Mode is not a new discipline with its own rulebook. It is SEO done well, on content good enough that an AI system has a reason to trust and cite it. The businesses that treat it that way, and resist the gimmicks, are the ones being referenced today, and they will be the ones still referenced as the systems get sharper.
That is the work we do. If you would like a clear view of where your site stands in AI search, and a plan to earn a place in the answers your customers are already reading, talk to our team or explore our digital marketing and SEO services.
