Google published a list of the specific things that help your content perform inside its AI experiences, AI Overviews and AI Mode. It is refreshingly free of tricks. There is no secret file, no AI-only markup, no rewrite aimed at machines. Instead there are eight moves, and every one of them is a fundamental you already know applied to a search surface that reads longer, more specific questions than the old box did. This post rewrites Google's post on performing well in AI experiences in plain English, walks each of the eight, and then splits them four ways so you can act whether you run an agency, a one-person shop, a growing team, or a company with many locations. If you want the companion piece on why the fundamentals are the whole game, read how to show up in Google AI Overviews and AI Mode.
The plain-English version
People ask AI search longer, more specific questions than they typed into the old search box. So the pages that win are the ones that answer a real, specific question well, on a site a machine can read fast. That is the whole idea. The eight moves below are just the parts of that idea broken out so you can check them off. None of them is new. What is new is that a shallow page which used to scrape a ranking now gets skipped, because the AI answer is looking for the page that actually resolves the question, not the page that mentions the keyword.
If you take one line away, take this: be the most genuinely useful, technically clean, honestly current page on your topic, and you are already doing AI search optimization. The eight moves are how you get there on purpose instead of by accident.
The eight moves, one at a time
1. Create unique, non-commodity content
This is the one that matters most, because AI search users ask harder questions. Someone typing a long, specific question does not want the same summary they could get from ten other pages. They want an answer with real expertise behind it. Recycled common knowledge gives an AI system no reason to pick you over anyone else, since every competitor has the same generic paragraph. What earns a citation is the thing only you can say: first-hand experience, real detail, a clear point of view, the answer you give customers on the phone that is not written down anywhere else.
- Write down the specific questions people actually ask you, in your own words, with your own experience behind the answer.
- Add the detail a generic page cannot: the edge case, the exception, the number from your own work (framed honestly as your own, not invented).
- Have a point of view. A page that takes a position is more citable than one that hedges everything.
- Skip the topics a hundred pages already cover the same way. Your version has to add something or it will not get pulled in.
2. Deliver an excellent page experience
Good content still loses if the page around it is a mess. A cluttered layout, a slow load, an interface that fights the reader on a phone, all of it undercuts content that would otherwise earn its place. Page experience is not a vanity metric here. It is part of whether the page is worth surfacing to a real person who will click through. Device matters and load time matters, and mobile is the default, not the afterthought. If your page takes five seconds to become usable on a mid-range phone, you have a page-experience problem no headline can fix.
- Test on a real phone on a normal connection, not just your fast laptop on office wifi.
- Cut the clutter that gets between the reader and the answer: interstitials, stacked pop-ups, layout that jumps as it loads.
- Treat load time as a feature. A fast page is a page people finish, and finishing is the engagement signal you want.
3. Make sure your content is technically accessible
If Google cannot crawl, index, and access a page, none of the rest counts. This is the price of entry. AI answers are grounded in pages that are actually in the index, so a page that returns the wrong status code, blocks the crawler, or hides its content behind rendering that machines miss simply is not eligible. Pages need to serve an HTTP 200 status and expose indexable material. It sounds basic because it is, and it is also the thing most often quietly broken: a robots rule that blocks more than intended, a noindex left on from staging, content that only appears after a JavaScript call the crawler never makes.
- Confirm your important pages return a 200 and are indexable, not accidentally noindexed or blocked in robots.txt.
- If your site is JavaScript-heavy, make sure the content is present without a click or a delayed render the crawler will not perform.
- Watch Search Console coverage for pages that dropped out of the index. Out of the index means out of AI answers, full stop.
This is the base layer everything else sits on. We wrote the checklist version in SEO foundations that make you eligible for Google AI search, and the eligibility mechanics in how pages get cited in Google AI search.
4. Use preview control tags thoughtfully
You have direct control over how much of your content can appear in a preview, and it is easy to strangle your own visibility without meaning to. The tags that govern this are nosnippet, the data-nosnippet attribute, max-snippet, and noindex. Each one restricts what Google can show. Set them too tight and you limit your own eligibility for AI features, because the system cannot use a passage you told it not to preview. There are real reasons to restrict a preview, a paywall, sensitive content, but restricting by accident or by blanket default is a self-inflicted wound.
- Audit for a stray nosnippet or an aggressive max-snippet limit on pages you actually want surfaced.
- Use data-nosnippet to fence off a specific element (a price you update hourly, a disclaimer), not the whole page.
- Reserve noindex for pages you genuinely do not want in Search. It also removes them from AI answers, so use it on purpose.
5. Align structured data with visible content
Structured data still helps. It makes your pages eligible for rich results and other features, and it gives machines a cleaner read of what a page is about. But there is a rule that trips people up: everything in your markup has to also be visible to a person on the page. Markup that describes content the user cannot actually see is a policy violation, not a shortcut. So structured data is a way to clarify what is already there, not a back channel for stuffing in claims, keywords, or a rating that appears nowhere on the page.
- Mark up what is genuinely on the page: the product, the FAQ answers, the article, the review that a visitor can read.
- Never describe in markup something a person cannot see. If it is worth telling a machine, put it on the page for the human too.
- Use structured data for the reasons it always existed, eligibility for features, not as an AI-only trick. There is no secret AI schema.
6. Enhance with high-quality images and video
Search is multimodal now. People search with a photo, not only with words, and answers increasingly pull in images and video where they help. Good visual assets do two jobs: they support understanding on the page, and they make you eligible to surface in image and video results and in visual-search answers. For products and local businesses there is an extra, concrete task attached to this: keep your Google Merchant Center feed and your Google Business Profile current, because that is how accurate prices, availability, hours, and location details surface inside AI answers about what to buy or where to go.
- Add relevant images and video that actually aid understanding, not stock filler. Give images descriptive alt text and file names.
- If you sell products, keep the Merchant Center feed accurate: prices, availability, and details drift, and stale data becomes a wrong AI answer.
- If you have a location, keep the Business Profile current: hours, address, services. A wrong hour is a wrong answer a system will confidently repeat.
7. Measure total visit quality, not just click volume
This is the one that changes how you read your own reports. Google says clicks from AI Overviews tend to be higher quality: the people who click after reading an AI answer spend more time on the page, because the answer already qualified them. So if you judge AI search purely on raw click count, you will underrate it and maybe abandon work that is quietly paying off. The number that matters is what happens after the click. Track conversions and engagement, not just how many clicks landed. A smaller number of better-qualified visits can beat a bigger number of bounces.
- Look past click count to what the visit does: time on page, pages per session, form fills, calls, purchases.
- Expect AI-driven clicks to convert differently. Fewer, warmer visits is a feature of this channel, not a failure.
- Set up the measurement before you judge the channel, so you are reading engagement and conversions, not just a traffic line.
Search Console is where you start reading AI-search visibility. We walk through it in measuring AI search visibility in Google Search Console so you can tell which pages earn impressions from AI-driven queries and double down on what works.
8. Adapt to evolving search behavior
The last move is a posture, not a task. Search keeps shifting, and it always has: desktop gave way to mobile, text picked up voice, and now voice and text share the stage with multimodal and visual search. Optimization is not a project you finish. It is a routine you keep, because what your users need and how they ask for it keeps changing. The good news is that the previous seven moves are exactly what keeps you ready for whatever comes next. A fast, clean, current, genuinely useful site adapts to a new search behavior far more easily than a brittle one built around one year's tactic.
The same posture positions you for what Google flags on the horizon: agentic experiences, where an AI system acts on a user's behalf rather than just answering. If your prices, hours, and offerings are current and machine-legible today, you are already positioned for the agent that reads them tomorrow. The framing to sell that strategy against is answer engine optimization.
What this means for you, by business type
The eight are universal. What changes is how you deliver them given your size, your team, and your appetite for risk. Here are the same eight, retold four ways.
If you run an agency
Turn the eight into a repeatable audit you can run across the whole client book. Make each one a line item with a pass or fail: unique expert content, page experience, technical accessibility, preview-tag hygiene, structured data that matches visible content, image and feed quality, measurement wired to conversions not clicks, and a standing cadence to re-run the whole thing. Sell the outcome, being eligible for citation in AI answers, not the mechanics. Deliver it as a scope a client understands: a crawl-and-index audit, a content-quality pass that adds real expertise, a tag and markup cleanup, and feed or profile hygiene for anyone with products or locations. Clients in professional services feel this most, and we built the audience framing at professional services.
The risk to manage at agency scale is move one and move five: never let volume push you into commodity content or markup that does not match the page. Standardize the checklist, not the words. This is where a platform earns its keep. Elevi runs the repeatable parts, the crawl checks, the tag audits, the feed monitoring, the visibility reporting, across the book so your people spend their time on the expertise that cannot be templatized. See how we structure delivery in our solutions.
If you are a micro business
You have no team, so you cannot run all eight at once, and you should not try. Two or three move the needle first. Start with move three: make sure your site is crawlable and returns a clean page that says clearly what you do, where, and for whom. Then move six: keep your Google Business Profile complete and current, because that is what lets AI answers about your kind of business in your area surface you at all. Those two are the base layer, and you can do both yourself between jobs without hiring anyone.
The third move, once those are solid, is move one: write down the answers only you know. The questions customers actually ask you on the phone, in your own words, with your own experience. One honest, specific page beats ten generic ones, because generic is exactly what an AI system has no reason to cite. Ignore anyone selling you a special AI file or AI markup; Google's own list does not include them. Just be findable, be current, and be specific. For a plain path sized for an owner-operator doing the work themselves, start at our page for micro businesses.
If you are an SME
You are past the point where one person holds everything, so the fix is to give each of the eight an owner and make it a quarterly routine. Assign it plainly: one person owns content quality (moves one and five), one owns the technical base (moves two, three, and four), one owns images, feeds, and profiles (move six), and one owns measurement (move seven). Move eight, adapting, is the reason you run the whole thing on a schedule instead of once. Quarterly is enough if you actually do it. Consistency beats intensity, and a steady drip of specific, well-structured content is exactly what AI search rewards.
The part SMEs skip and should not is move seven, measurement. You cannot manage what you do not watch, so wire up conversions and engagement, not raw clicks, and learn to read AI-search visibility in Search Console. The routine plus named owners plus real measurement is a system you can run without a big team. Our sizing and the offer that fits it live at our page for small businesses. If you are at the top of this band and starting to add locations or product lines, the governance in the next rung, mid-size, is where these owners turn into standards.
If you are a mid-size company
Your problem is not knowing the eight. It is running them consistently across many teams, locations, and pages without brand drift or a compliance slip. So operationalize them. Turn moves one through five into a published content-and-technical standard that is the definition of done for every team that ships a page: what unique and expert means, the page-experience bar, the crawlability and status-code checks, the preview-tag policy, and the rule that markup must match visible content. Centralize the technical fundamentals so a hundred contributors cannot each reinvent them badly, and make move six, feed and profile hygiene, a governed job across every location, because one stale price or hour is a wrong answer an AI system will repeat with confidence.
Then wire reporting to move seven at the org level: track conversions and engagement by team and by location, not a single vanity click line, so you can see which units are actually earning AI visibility. Move eight becomes a review gate, not a hope: a standing owner watches how search behavior shifts and updates the standard, so no eager team quietly ships commodity content or mismatched markup. Do this well and when agentic experiences arrive, your fast, structured, current site is already positioned across every location at once. Our framing at this scale is our page for large companies, and if you span many brands or business units, enterprise is the bracket built for that multi-brand governance problem.
Common questions
Which of the eight matters most?
Move one, unique and expert content, because AI search users ask longer, more specific questions that need real expertise to answer. But it only pays off on top of move three, technical accessibility. A brilliant page Google cannot crawl or index earns nothing. Get crawlable and indexable first, then make the content the best answer on your topic.
Do preview tags really hurt my AI visibility?
They can, if they are set too tight. nosnippet, an aggressive max-snippet limit, a stray noindex, or a blanket data-nosnippet all restrict what Google can preview, and a passage it cannot preview it cannot use in an AI feature. Use them on purpose for a real reason, and audit for a template or plugin that set one for you by default.
Should I judge AI search by clicks?
No. Google says clicks from AI Overviews tend to be higher quality, with visitors spending more time on the page. If you measure only raw clicks you will underrate the channel. Track total visit quality, conversions and engagement, so a smaller number of better-qualified visits gets the credit it earns.
Do I need a special file or AI-only markup for any of this?
No. None of the eight is a special file or AI-only markup. Structured data helps, but only when it matches what a person can see on the page, and it is the same structured data that always earned rich results. If a vendor's pitch is built on a secret AI file, check it against Google's own guidance first.
The honest headline is that these eight are not a new game. They are the fundamentals, done well and on purpose, on a site machines can read. If you want to know where you stand and which of the eight to fix first, run the estimator for a plain read on your site, or talk to us and we will point you at the move that will do the most for your size. Not sure which bracket you are in? Find yourself on the who-we-serve pages and start from the version written for a team like yours.
