How to Optimise for Google’s AI Overviews (AIOs)

Optimising for Google’s AI Overviews (AIOs) is mostly a matter of doing SEO well. Google says as much in its official guide to generative AI search, which describes the work as optimising for the search experience, “and thus still SEO.” That does not mean nothing has changed. AI Overviews and AI Mode now answer more searches on the results page, clicks are falling for many sites, and some popular tactics have turned out to be unnecessary. This guide covers what AIOs are, what Google actually asks of site owners, what to do about it and how to tell whether it is working.

What are AI Overviews, and how is AI Mode different?

AI Overviews are AI-generated summaries that appear at the top of some Google results. They pull together information from several web pages and show links to the sources underneath.

AI Mode is a separate, conversational way to search. Instead of one summary above a list of results, you ask a longer question, follow it up with more, and Google answers each one with links to supporting pages. In the UK, AI Overviews arrived in 2024 and AI Mode followed in July 2025, where it appears as a tab on the results page.

Both draw on the same index as ordinary search. Google describes two techniques behind them. The first retrieves relevant, current pages from the Search index and uses them to ground the answer. The second, called query fan-out, runs several related searches at once to gather more material. Someone asking how to start a herb garden on a balcony, for example, might trigger side searches about herbs that cope with shade, pot sizes and watering. That is a good reason to cover the obvious follow-up questions on a topic within the page itself.

How AI Overviews affect traffic

The effect on clicks is measurable. Ahrefs studied 300,000 keywords, half with an AI Overview and half without, using December 2025 data. It found that an AI Overview goes hand in hand with about 58% fewer clicks for the top organic result. The same study run in April 2025 found a 34.5% drop. Lower positions lose clicks too, with the fall shrinking to about 19% at position ten. You can read the method in Ahrefs’ update on AI Overviews and clicks.

Treat that as a strong signal rather than a verdict. It is a correlation across a large sample of informational keywords, not a measurement of your own site. What it does tell you is that ranking alone no longer guarantees the visit. Being one of the sources an AI Overview cites matters, and so does giving people a reason to click anyway. Our post on zero-click searches looks at that second half in more detail.

What Google says you need to do

In May 2026 Google published its first official guide to optimising your website for generative AI features on Google Search, and it is direct about what is required. A page has to be indexed and eligible to appear in Google Search with a snippet. The site also has to be included in Search generative AI features in Search Console. Beyond that, Google lists no extra requirements, and meeting them does not guarantee that a page will be crawled, indexed or shown.

The same guide names several tactics you can ignore for Google Search:

• llms.txt files and other special AI files, markup or Markdown versions of your pages. Google Search does not use them. Creating them for other services is fine, but it will not help or harm your visibility in Google.

• Breaking content into tiny chunks. Google can handle several topics on one page, and there is no ideal page length.

• Rewriting pages in a special style for AI systems. Google’s systems understand synonyms and meaning, so you do not need to cover every long-tail variation of a question.

• Chasing inauthentic mentions across the web. Google’s systems are built to favour quality and to filter out spam.

• Treating structured data as a requirement. It is not needed for generative AI search and there is no special schema to add, although it still helps with rich results.

That corrects some of the advice that circulated when AI Overviews were still called Search Generative Experience (SGE), including the idea that special schema or a rigid page format will earn you a citation.

What to do instead

Publish content only you could have written

Google puts this first. It says unique, useful content will likely do more for your presence in generative AI search than any other suggestion in the guide, and it contrasts “non-commodity” content with pages that restate what is already online. A list of seven generic tips is commodity content. A write-up of what you tried, what it cost and what happened is not.

In practice that means:

• original data, even from a small sample, with the method explained

• case studies and worked examples from real projects

• named authors with relevant experience and a short bio

• a clear point of view where one is warranted, with the reasoning behind it

The difference shows up even at paragraph level. A thin version reads: “Link building is good for SEO.” A stronger one reads: “Links from relevant, reputable sites show Google that other publishers consider a page worth referencing, which can help it rank. A link from a trade publication in your own sector usually does more for you than ten from unrelated directories.” The second version answers the question and adds context a reader can use. Our post on first-hand experience in content covers how to bring that experience into a page.

E-E-A-T works well as a checklist here. Google’s quality raters look for Experience, Expertise, Authoritativeness and Trustworthiness, so show each one on the page: first-hand knowledge of the subject, author credentials, a reputation that others can confirm, and basics such as a secure site and clear authorship.

Answer the question early, then add depth

Open each section with a direct answer, then explain it. Use headings that match the questions people ask, numbered lists for processes and bullets for options. Google says clear headings, sections and paragraphs help readers, and that there is no need to split pages into fragments for AI.

AI Overviews show up most often on informational searches, such as how-to questions, explanations and comparisons, so these are the formats where clear structure pays off. For example, a section headed “What is a canonical tag?” could begin: “A canonical tag tells search engines which version of a page you want indexed when the same content appears at more than one URL.” The sentences that follow can explain when to use one and what goes wrong without it.

Then cover what a reader would ask next. Google’s systems run related searches behind the scenes, so a page that deals with the obvious follow-ups has more to offer. Stay on one topic, though. Creating a separate page for every variation of a question, mainly to influence rankings or AI answers, breaks Google’s scaled content abuse policy, and a large number of pages does not make a site better.

Get the technical basics right

AI features use the same crawling and indexing as normal search, so the standard checks apply:

• Make sure robots.txt, and any CDN or firewall, lets Googlebot in.

• Keep important content in text rather than locked inside images.

• Link to the page from other relevant pages on your site.

• Check that pages load quickly and read well on a phone.

• Keep duplicate versions of the same content to a minimum.

• Confirm in Search Console that your site is included in Search generative AI features.

Local businesses and online shops can also appear in AI responses through Google Business Profile and Merchant Center, so keep those details accurate.

Add images and video where they help

Google’s AI features can bring in relevant images and video, which gives your pages more ways to appear than a text link alone. Use original, high-quality visuals that support the text, with descriptive file names and alt text, as you would for image SEO.

If clicks fall, plan for it

Some searches will be answered on the results page whatever you do, so build pages that give people a reason to visit. Calculators, templates, original research and tools offer something a summary cannot. Invite readers to subscribe or download something useful early in the visit, because they may only come once. And keep building the brand, because people who already know you can come straight to you.

How to see whether it is working

Start with the searches you already care about. In Ahrefs Site Explorer, open the Organic keywords report and use the SERP features filter to select AI Overview. That shows which of your rankings trigger one. Then run those searches yourself, including the AI Mode tab, and see which sources are cited.

For your own visibility, use Search Console. On 3 June 2026 Google launched Search Generative AI performance reports, with separate views for Search (AI Overviews and AI Mode) and Discover. They show impressions in those features by page, country, device and date. At launch they covered impressions rather than clicks, so they show where you appear, not how many people visited from there. Google introduced them to a subset of sites and widened access over the summer. Our post on what Google’s new AI search reports mean for your SEO strategy explains how to read them.

Look at which pages appear most often. If a thin or outdated page is the one Google keeps surfacing, improve it before you chase impressions elsewhere. Be careful with third-party tools that claim to show Google’s internal AI metrics. Google says no outside tool has access to its ranking or AI systems, so treat their numbers as estimates.

Getting help with original content

If your team does not have the time to produce expert-led content at the level AI features reward, our content writing service can help you plan and write it.

The short version

Make sure your pages can be indexed, publish things worth citing, answer questions plainly and measure with Search Console rather than guesswork. Skip the special files, the chunking and the schema hunt. Google’s guide is clear that none of them is needed, and the time is better spent on the content itself.