Google AI Overviews Optimization: What Google Says and What to Do

To appear in Google AI Overviews and AI Mode, do what Google already asks for in Search: have an indexed, snippet-eligible page with unique, helpful content, keep your site included in generative AI features in Search Console, and measure with the new Generative AI performance report. Google says there are no special files or markup to add.

This is the shortest honest answer, and it comes straight from Google. In 2026 Google published a dedicated guide to generative AI features, a new Search Console setting and a new report, so for the first time there is official material to build on rather than guesses. This article walks through each piece in order, explains what you can and cannot control, and sets out a practical plan. It covers only Google. For ChatGPT and Perplexity, see our guides to ChatGPT search and Perplexity.

AI Overviews and AI Mode, explained

Google AI Overviews are AI-generated summaries that appear on some search results pages, with links to supporting pages. Google AI Mode is a more conversational experience for questions that need exploration, reasoning or comparison. Google reported in June 2026 that AI Overviews had more than 2.5 billion monthly users and AI Mode more than one billion. Both features are described by Google as part of Search, not as a separate product with its own rules.

	AI Overviews	AI Mode
Where it appears	On some standard results pages	In a dedicated, conversational Search experience
Best suited to	Getting the gist of a topic quickly	Complex questions, comparisons and follow-ups
Links to sites	Yes, as supporting links	Yes, as supporting links
Counted in Search Console	Yes, in the Generative AI performance report	Yes, in the Generative AI performance report

What Google says in its official guide

Google's guide, Optimizing your website for generative AI features on Google Search, was last updated on 10 July 2026. Its main messages are these.

• SEO still applies. Google says its generative AI features are rooted in its core ranking and quality systems, so SEO best practices stay relevant. It treats AEO and GEO as labels for work on these experiences, and says that from its side, this is still SEO.
• Two techniques sit underneath. Retrieval-augmented generation grounds answers in pages retrieved from the Search index. Query fan-out runs several related searches at once. Google gives the example of a question about fixing a weedy lawn fanning out into searches about herbicides, chemical-free removal and prevention.
• Unique, non-commodity content matters most. Google says this will likely influence your presence more than any other suggestion. Commodity content restates common knowledge, while non-commodity content carries an expert or first-hand point of view.
• Technical clarity. Pages must be crawlable, indexed and snippet-eligible. JavaScript content can be processed as long as it is not blocked, but Google notes that JavaScript sites are generally more complex to get right.
• Page experience and media. Pages should work on all devices, load quickly and make the main content easy to find. High-quality images and video give more ways to appear.
• Business details. For products and local businesses, Merchant Center feeds and Google Business Profiles help your details appear in AI responses and elsewhere in Search.

The things Google says you can ignore

The guide has a list of popular tactics that Google says are not needed for Google Search. This is the most useful part for anyone being sold "AI optimization".

Tactic	What Google says	Our reading
llms.txt and other special files or markup	Google Search does not use them. Creating them neither helps nor harms visibility	Optional for other systems. Not a Google fix
Breaking content into tiny chunks	No requirement. Google can find the relevant piece on a multi-topic page. No ideal page length	Write for readers. Do not split pages for AI
Rewriting content just for AI	Not needed. Systems understand synonyms and meaning, so you need not capture every long-tail phrase	Edit for clarity, not for a model
Seeking inauthentic mentions	Not as helpful as it seems. Core ranking systems favour quality, and spam systems block spam	Earn real mentions. Never buy them
Over-focusing on structured data	Not required, and no special schema.org markup exists for this. Still worth using for rich results	Use it where it matches visible content

Google also warns against mass-producing separate pages for every query variation mainly to manipulate results. It says doing so violates its scaled content abuse policy and is ineffective over time. This matters for anyone planning a large content programme, ours included. Every page needs its own distinct purpose and its own search intent, which is why we map one primary keyword to one URL.

The new Search Console control

On 3 June 2026 Google announced a control in Search Console that lets site owners decide whether their site appears in, and helps ground responses in, generative AI features such as AI Overviews, AI Mode and AI Overviews in Discover. Google says sites that opt out will not receive traffic or impressions from those features, and that the control will not be used as a ranking signal for results outside them. Google's guide adds that a site must be included in these features to be eligible to display there. As of 31 August 2026, Google says the features have rolled out to all websites worldwide.

Two practical points follow.

1. Check the setting. If your site is excluded, whether by a past decision or by a colleague, no amount of content work will bring you into these features.
2. Treat opting out as a business decision. Some publishers choose it. For most businesses that depend on discovery, staying included is the default.

Source: Google, New opportunities, control and insights for website owners.

Measure with the Generative AI performance report

Search Console now includes a Generative AI performance report for Search, and a separate one for Discover. Google's help page says it shows how your organic impressions in generative AI features change over time, which pages get the most and least impressions, and where those impressions come from by device or country. It currently covers AI Overviews and AI Mode.

You can see	You cannot see (today)
Impressions in AI Overviews and AI Mode	Clicks from those features, on their own
Which pages appeared	Average position or click-through rate in those features
Country and device splits	The exact queries that triggered each impression
Trends by day, week and month	Data from Search Labs experiments

Google says the same data also feeds the overall performance report, and that it plans to add metrics over time. Check the help page for the current list, as it may have changed. Source: Generative AI performance report (Search) in Search Console Help.

How to use it well:

• Compare pages with high AI impressions against pages with high clicks. The overlap and the gaps tell you which pages work as a source and which as a destination.
• Watch for pages with many AI impressions and falling clicks. They may be satisfying people in the answer. Decide whether to add something worth clicking for.
• Do not read AI impressions as a ranking position. They are a different measure.
• Be wary of third-party tools that claim internal Google metrics. Google says no third party has access to its internal ranking or AI systems.

A practical plan for AI Overviews and AI Mode

1. Confirm eligibility

Check that your site is included in generative AI features in Search Console. Confirm that key pages are indexed, return 200, are not set to noindex and are not blocked by nosnippet or a very low max-snippet setting. Google states that more restrictive preview controls limit how your content can appear. Check robots.txt for Googlebot, and your CDN and firewall.

2. Find the pages that matter

List the pages that drive enquiries and sales. Open the Generative AI report and see which already appear. Start with the important ones that do not.

3. Raise the uniqueness of each page

Ask what your page offers that a summary of existing pages does not. First-hand experience, original data, a worked example, a real photo of the process, a genuine expert's view. Remove or merge pages that only restate common knowledge.

4. Cover the neighbouring questions

Because of query fan-out, a topic is investigated through several related searches. Make sure your page, or your cluster of pages, answers the obvious neighbouring questions. Google also cautions against creating separate pages for every variation just to manipulate results. The test is whether a visitor would find the page useful on its own.

5. Write clearly for people

Use clear headings, short paragraphs and direct answers where a reader would want one. Google does not require a special style for AI, and we recommend this because it helps readers.

6. Add media that helps

Support the text with relevant, high-quality images and video, following Google's image and video SEO guidance. Add descriptive alt text and sensible file names.

7. Keep business data accurate

For shops, keep Merchant Center feeds accurate. For local businesses, keep your Google Business Profile complete and current, with correct services, hours and contact details.

8. Use preview controls deliberately

The controls that govern snippets, nosnippet, data-nosnippet and max-snippet, also affect how your content can feature. Use them only where you have a reason to limit display.

9. Monitor, then refresh

Review the Generative AI report monthly, update pages whose facts changed and note what you changed so you can learn from the results.

Reading the report: five scenarios

Numbers only help if you know what to do with them. These are the patterns we see most often, with a sensible response to each. They are illustrations of how to reason, not statistics.

Pattern	What it may mean	What to do
High AI impressions, healthy clicks	The page works as both a source and a destination	Protect it. Update it on schedule and avoid risky rewrites
High AI impressions, falling clicks	The answer may satisfy people without a visit	Add something worth clicking for, such as a tool, data or a worked example
Few AI impressions, strong classic rankings	The page ranks but is not chosen as a supporting source	Compare with the sources that are included. Look for unique value you lack
No AI impressions on important pages	An eligibility issue, thin content or low relevance	Check indexing, snippet settings and the Search Console AI setting first
AI impressions rising on a page after an update	The change may have helped	Note what changed and apply the lesson to similar pages

Commodity vs non-commodity: examples for different businesses

Google's guide uses a homebuyer example to explain commodity and non-commodity content. Here are invented examples in other industries to make the idea concrete.

Business	Commodity (easy to replace)	Non-commodity (hard to replace)
Accountancy	"10 tips for filing your tax return"	"What we saw when 40 sole traders filed late: a breakdown of the most common causes, anonymised"
Dental practice	"What to know about dental implants"	"A clinician's walkthrough of our implant process, with recovery photos from consenting patients and our own complication rates"
Web agency	"Why site speed matters"	"How we cut a client's mobile load time, step by step, with before and after measurements and what did not work"
Online shop	"How to choose running shoes"	"Fit notes from our returns data: which sizes run small in which models and why"
Software company	"What is project management software?"	"How three customers migrated from spreadsheets: timelines, pitfalls and the template we built"

The common thread is first-hand evidence or an original viewpoint. Use only real experience, with consent where people are involved, and do not invent examples to look authoritative.

An image and video checklist

Google says high-quality images and video can bring more opportunities to appear in generative AI features, and points site owners to its image and video SEO guidance. A practical checklist:

• Use original photos and diagrams where you can, not generic stock that appears on thousands of sites.
• Give each image a descriptive file name and an alt text that describes what it shows.
• Place images near the text they support, and keep the text as real text, not inside the image.
• Compress images and use modern formats so pages stay fast.
• For video, host it on a page that also contains a written summary or transcript, and add structured data only where it matches the visible content.
• Add captions that explain why the image matters.

A 60-day schedule

Days	Work	Output
1 to 7	Eligibility check: Search Console AI setting, indexing, snippet controls, robots.txt, firewall	A list of issues, fixed
8 to 21	Review the Generative AI report, classify pages using the scenarios table, choose the ten that matter	A prioritised page list
22 to 45	Upgrade the top pages with original evidence, clearer structure, media and accurate business data	Improved pages with truthful update dates
46 to 60	Re-check the report, compare clicks and impressions, note lessons	A short review and a plan for the next 60 days

How AI Overviews relate to featured snippets

Featured snippets and AI Overviews both put an answer at the top of the page, but they work differently. Google says it decides whether a page would make a good featured snippet, and you cannot mark a page to become one. AI Overviews instead draw on several sources and write a new summary, with links to supporting pages. A page can win one without the other. Writing a clear, direct answer helps readers in both cases, and our AEO guide covers the answer-first craft. Do not assume that holding a featured snippet means your page will be linked in an AI Overview, or the reverse.

A note on Google Discover

Google also reports generative AI features in Discover, and Search Console has a separate Generative AI performance report for Discover. If a large share of your traffic comes from Discover, for example if you publish news or lifestyle content, check that report separately. The Search Console setting that governs AI features covers AI Overviews in Discover as well, according to Google's announcement.

Agentic experiences: a note on what is next

Google's guide also points to agentic experiences, where AI agents act for a person, such as comparing products or booking something. It says browser agents may read a site through screenshots, the page structure and the accessibility tree, and it refers to guidance on making sites agent-friendly and to emerging protocols for agent commerce. If your business sells online or takes bookings, make your key flows work with plain HTML, clear labels and accessible forms. These improve usability for people too. We explain the wider trend in what AI agents actually are.

Google-Extended and other controls

Google's AI features in Search are tied to normal Search access, so robots.txt rules for Googlebot are the control for crawling, and snippet controls and noindex limit display. Google says its separate Google-Extended control relates to limiting AI training and grounding in some of Google's other systems. It is a different switch from the Search Console setting for AI features in Search. Review Google's list of common crawlers before changing robots.txt, and test any change.

Local businesses and online shops

Google says generative AI responses can include product listings, product information and information about local businesses. Where that applies to you:

• Online shops: keep product titles, prices, availability and shipping details accurate in your feeds and on the page. Write buyer-focused answers on sizing, compatibility, delivery and returns.
• Local firms: complete your Business Profile, keep details consistent, respond to reviews and publish genuinely useful pages for the areas you serve.

If you want hands-on help with the foundation, our local SEO services and e-commerce SEO services cover these areas.

Common mistakes

• Paying for an "AI Overview optimization" package built on llms.txt, chunking or markup tricks that Google says are not needed.
• Ignoring the Search Console setting, so the site is excluded without anyone knowing.
• Publishing dozens of near-duplicate pages for question variants.
• Judging results by clicks alone and missing impressions in the new report, or the reverse.
• Rewriting a page that already ranks and breaking what worked.
• Blocking snippets by accident with a restrictive meta tag.
• Believing a vendor who claims inside knowledge of Google's AI ranking.

Where FAQ content and schema fit

Google says structured data is not required for generative AI search and that no special schema.org markup exists for it, though it remains useful for rich result eligibility. FAQ blocks are useful for readers with a specific doubt, and for the wider history of FAQ markup in Search see our note on Google removing FAQ rich results. Use markup only where it matches visible content.

Quick glossary

Term	Meaning
AI Overview	An AI-generated summary on some Google results pages, with links to supporting pages
AI Mode	A conversational Google Search experience for complex questions
Grounding	Basing an AI response on retrieved web pages to improve accuracy and freshness
Query fan-out	Running several related searches at once to build an answer
Non-commodity content	Content with a unique expert or first-hand point of view
Preview controls	nosnippet, data-nosnippet, max-snippet and noindex settings that limit display

Where we can help

We are a digital marketing agency in Manchester, UK and Mumbai, India. Our AI SEO services begin with an eligibility and Generative AI report review, then focus on the pages that drive revenue. Contact us for a short review, and we will tell you plainly where the opportunities are and which ideas are not worth your money.