BlogGuides
How to rank in AI Overviews: Google is placing more links inside its AI answers, not just beneath them
On 15 October 2026 Google said it is adding more inline links to AI Overviews and AI Mode, showing them higher, and testing a top-of-answer section for official pages. Being the page an answer links is now a contest decided before the organic position you already fight for.
This post reports what Google said and what it did not, and then does the part only a scanner for AI visibility can add: it walks through what has to be true of a page before any of these surfaces can link it. None of the numbers here are ours. The one figure Google published, and the status of each change, come from that announcement and are dated to it, because a claim about another company's product is a claim about the day you read it.
In short
- Google announced on 15 October 2026, in a post by vice president of product management Hema Budaraju, that it is making substantial increases to the presence of inline links in AI Overviews and AI Mode and showing those links higher in responses more often, so the link placed inside an AI answer is now the thing a page competes for.
- How to rank in AI Overviews now means being the page Google links as a source inside the answer, not only the page it ranks beneath it: the organic position and the cited link are decided separately, so a page can hold one without the other.
- Google said on 15 October 2026 that the number of people using its Preferred Sources feature has grown by more than 7x since the start of the year, which is a count of people choosing sources and not a measure of referral traffic to those sources.
- A dedicated section at the top of some AI answers for official pages, and clickable creator and publisher names on video results in AI Mode, were described as tests on 15 October 2026 rather than shipped features, so neither is a placement a page can yet be optimised for.
- A page can only be linked inside an AI answer if a crawler could fetch and read it first, so eligibility rests on the same crawl access, server rendered prose and structured data that decide every other AI surface, measured on the day you check because all of it can change.
| Change | Where it appears | Status Google gave it |
|---|---|---|
| More inline links in answers | AI Overviews and AI Mode | Rolling out |
| Inline links shown higher, more often | AI Overviews and AI Mode | Rolling out |
| A section for official pages | Top of some AI responses | Test |
| Timely-news carousels of Preferred Sources | More locations and languages | Rolling out |
| Larger related-article links with thumbnails | End of the answer | Rolling out |
| Clickable creator and publisher names | Video results in AI Mode | Test |
What did Google change in AI Overviews and AI Mode?
Google grouped five changes under one idea: more ways out of the answer and back to the open web. The first two are the load-bearing ones. Google said it is substantially increasing how many inline links appear in AI Overviews and AI Mode, and showing those links higher in the response more often. It attached no percentage to substantially, so there is no figure to quote here, only a direction of travel, and this post does not invent one to fill the gap. This blog has already pulled apart the announcement's own numbers, noting that the one figure it carries is a multiple with no base, and the question here is the next one: what a page has to be before any of these new links can land on it.
The other three are smaller and two of them are explicitly tests. Google said it is testing a dedicated section at the top of some AI responses that points to official pages, with a short note on why each link is useful. It said end-of-answer related-article links are getting image thumbnails and a more clickable design. And it said it has started testing clickable creator and publisher names on video results in AI Mode, each linking to a Search profile. A test is not a placement a page can be optimised for, and treating one as shipped is how a guide ages badly inside a week, so the hero table above marks each change with the status Google itself gave it.
The honest read is that this is Google responding to a complaint, not Google publishing a ranking signal. None of it tells you how a page is chosen to be one of those links. For that you have to look at what an answer engine does before it writes a word, which is the same pipeline behind how to get cited in Google AI Overviews and behind being named by any other engine. The link surface got wider. The bar to be on it did not move, and the rest of this post is about where that bar actually sits.
The older shape
- The answer is generated at the top
- Source links sit beneath it, easy to skip
- Being ranked and being read are close together
What Google is expanding
- More links placed inside the answer itself
- Links shown higher in the response more often
- Being the linked source becomes its own contest
How do you rank in AI Overviews when ranking now means being the linked source?
Start by dropping the word rank for a second, because it carries a decade of blue-link assumptions that do not transfer cleanly. An AI Overview does not sort ten pages and show them in order. It assembles an answer and attaches links to the sources it leaned on, and the change Google announced is that it will attach more of them and place them more prominently. So the target is not a position, it is being selected as a source the answer links. That is why this blog separates the two outcomes the way the crawl GEO and SEO share does: the same fetch feeds both, and everything after the fetch is decided differently.
Being selected as a source has a precondition that no amount of on-page polish replaces. The page has to be reachable and readable by the crawler that feeds the answer, because a page that cannot be fetched is not a weak candidate, it is not a candidate at all. This is the step most ranking advice skips, and it is the step an AI crawler either clears or does not. Once a crawler has the HTML, the next question is whether the words a reader sees are the words the crawler got, which is prose parity, and the one after that is whether a single passage cleanly answers the query, because a passage is the unit an answer lifts.
Then, and only then, do the softer signals matter: whether the page carries structured data that matches what it says, and whether an engine can tell who is answering well enough to attribute the source. The order is the point. A site can spend months on schema and tone and win nothing, because the crawler was turned away at the first step and never saw any of it. The same logic governs getting cited by ChatGPT and getting cited by Perplexity: the surfaces differ, the precondition does not.
Flow: Query reaches AI Mode to Google retrieves candidates; Google retrieves candidates to Crawler can fetch the page; Crawler can fetch the page (refused) to Blocked, not a candidate; Crawler can fetch the page (fetched) to Prose readable in the HTML; Prose readable in the HTML to A passage answers the query; A passage answers the query to Linked inside the answer.
Can Google read the page it would link?
This is the question the announcement does not answer for you, and it is the one that decides everything downstream. A link inside an AI answer is a link to a page Google could fetch, parse and quote. If the page refused the crawler, or served it a shell that a browser later fills in with JavaScript, there is nothing to quote and nothing to link. The controls that govern this are old and documented, and Google ties two of them to the AI surfaces by name. Google's robots meta tag documentation states that nosnippet will prevent the content being used as a direct input for AI Overviews and AI Mode, and that max-snippet limits how much of the content may be used as a direct input. This blog counted who actually sends those directives when it looked at the snippet directive that governs AI Mode, and the answer was almost nobody on purpose.
There is a trap one level up, at the crawler itself. A site that wants to keep training data out often reaches for a robots.txt block, and Google-Extended is the token people name. Google's crawler documentation scopes Google-Extended to Gemini model training and grounding, and says plainly that it does not affect a site's inclusion in Google Search. AI Mode and AI Overviews are Search surfaces, so blocking Google-Extended does not remove you from them, and blocking Googlebot in the hope of blocking AI does remove you from both. The group-selection rules that decide which block applies are not Google's invention either: they are specified in the Robots Exclusion Protocol, RFC 9309, and a misread of them is one of the most common ways a site quietly disappears from a surface it meant to stay on.
The practical consequence is that the first thing to audit is not your schema or your headings. It is whether the exact page you want linked returns a readable response to an identified crawler, and whether the words are in that response rather than painted on afterwards. You can watch this happen on a single URL with the free check for what a crawler sees on your page, and the broader pattern across real sites is in our crawlability study.
| Control | What Google's documentation says it does | Where it lives |
|---|---|---|
| nosnippet | Prevents the content being used as a direct input for AI Overviews and AI Mode | Robots meta tag or X-Robots-Tag header |
| max-snippet | Limits how much of the content may be used as a direct input | Robots meta tag or X-Robots-Tag header |
| robots.txt Disallow | Stops the page being crawled, so it can never be read or linked | robots.txt |
| Google-Extended | Governs Gemini training and grounding, not inclusion in Search | robots.txt user-agent token |
What are Preferred Sources, and does marking one help?
Two of the five changes lean on a feature called Preferred Sources, so it is worth being exact about what it is. Preferred Sources is a reader-side control: a person tells Google which sites they want to see more of in Search and its AI experiences, and Google weights those sites up for that person. Google said in the announcement that the number of people using the feature has grown by more than 7x since the start of the year. That figure is the one hard number in the post, and it is easy to misread, so to be precise: it counts people who have turned the feature on, not clicks, not sessions, and not referral traffic to any site. A reader choosing you is not the same as Google sending you traffic, and the announcement does not claim it is.
What changed for Preferred Sources is placement. Google said it is expanding timely-news carousels that highlight Preferred Sources to more locations and languages, without naming which, and that is a surface a publisher can be surfaced in rather than one it can switch on. The honest framing is that you cannot mark your own site as a Preferred Source; your readers do. What you can do is be the kind of source a reader would choose and a carousel would surface, which is a brand-and-content question more than a technical one, and it overlaps with what AI already says about your brand. This blog looked at the mechanics of the feature when it covered how Preferred Sources actually works, and the short version is that the button lives in the reader's settings, not in your markup.
The top-of-answer section for official pages is the other Preferred-adjacent change, and it is the one to be most careful about, because it is a test. Google described it as something it will be testing in some responses, with context on why each link is useful. Until it ships broadly, there is no documented way to qualify for it, and a guide that tells you how to win a placement Google is still testing is selling certainty that does not exist. Treat it as a signal of direction, not a tactic.
-
Preferred Sources carouselLive, expanding Reader chooses the sources; Google said usage grew more than 7x since the start of 2026 -
Top-of-answer official pagesTest A test in some responses, with a note on why each link is useful; not a shipped placement -
End-of-answer related linksLive, larger Now carries image thumbnails and a more clickable design -
Creator names on videoTest Clickable names linking to Search profiles, under test in AI Mode
Does structured data decide whether an answer links you?
It helps, and it is not the lever most guides make it. Structured data is how a page tells a machine what it is in a vocabulary the machine already knows, the schema.org vocabulary, and a page that labels its parts is easier to parse into the passage an answer lifts. But labelling is not a ranking cheat, and this blog has published the inconvenient version of that: when it ran a citation audit across more than a thousand URLs, structured data landed third, behind whether the content was reachable and whether it answered the question. Schema earns its place after those two, not instead of them.
The failure mode worth naming is schema that disagrees with the page. A product markup that lists a price the visible text does not show, an FAQ block that answers a question the page never asks, an author field naming someone who appears nowhere else: each of these is a signal an engine has learned to distrust, because it is the signature of markup bolted on for machines rather than describing the page a reader sees. The rule that keeps schema honest is that every value in it should appear in the words a person reads, which is the same discipline behind prose parity one level down. Markup that matches the page strengthens attribution. Markup that contradicts it weakens it.
So the structured-data answer to how an answer picks a source is: use it to describe, not to decorate, and only after the page is reachable and readable. A site that gets the order wrong spends its effort on the signal that matters least while the one that matters most, crawl access, is still failing silently. You can see which types your pages actually carry, and whether they match the text, in the same scan that checks everything above, and the method Lantad uses to weigh them is documented on the methodology page.
Sample Illustrative, not a measurement of any real site.
- Crawlable A crawler can fetch the URL: no block in robots.txt, no 403 to an identified bot
- Readable without JavaScript The answer text is in the served HTML, not injected by a script the crawler does not run
- One clear passage per question A heading states the question and the next lines answer it, so a passage can be lifted whole
- Structured data that matches the page schema.org types whose values also appear in the visible text, describing not decorating
- An unambiguous entity The page makes clear who is answering, so the source can be attributed with confidence
What should you check on your own site this week?
Work the order the pipeline uses, because a failure early makes every later effort invisible. First, confirm an AI crawler can fetch the page you want linked. The free check for which AI crawlers can reach your site and the robots.txt tester answer that from the outside, the way an engine sees it, and they will tell you if a Google-Extended block or a blanket rule is quietly taking Googlebot with it. Second, confirm the words are in the served HTML and not painted on by a script, because a crawler reads the response, not the rendered view. Third, read your own page as a list of question-and-answer blocks and ask whether any single passage could be lifted to answer a query, which is the test behind both how to get cited in Google AI Overviews and getting named across every answer engine.
Only then is it worth auditing schema and entity clarity, and the honest expectation is modest: these sharpen attribution, they do not manufacture a citation for a page an engine could not read. Nothing on this list is specific to Google. The announcement that prompted this post is Google adding links to its own answers, but the page that earns one of those links is the same page that earns a mention in ChatGPT or Perplexity, because all of them start by fetching the HTML and reading what is there.
That is the whole of the Lantad argument, and it is why the product exists. Paste a URL and Lantad reads your page as each of the crawlers that feed these answers, reports exactly what each one received and where it failed, and refuses to print a grade where it could not measure. If you want to know whether you can be the source Google now wants to link, the fastest answer is to scan your own URL and see what the crawler sees before an AI answer decides for you.
| Check | Why it decides AI linking | How to see it |
|---|---|---|
| Can an AI crawler fetch the page | A blocked page is never a candidate to be linked | Request it as the bot and read the status code |
| Does the served HTML carry the answer | A crawler reads the HTML, not the rendered view | Compare the crawler view with the browser view |
| Is there one liftable passage per question | A passage is the unit an answer quotes | Read the page as question-and-answer blocks |
| Does structured data match the words | Schema that contradicts the page is distrusted | Check each type against the visible text |
Lantad
Published .
The question people type has caught up with the product. How to rank in AI Overviews is now asked the way ranking in Google was asked a decade ago, and on 15 October 2026 Google changed what the answer has to account for. In Google's announcement, written by vice president of product management Hema Budaraju, Google said it is making substantial increases to the presence of inline links in its AI experiences and showing those links higher in responses more often. The sentence is short and the shift inside it is not: a link placed inside an AI answer, rather than listed under it, is the thing a page now competes for, and that is a different contest from the one generative engine optimization has been describing.
Common questions
How do you rank in AI Overviews?
Google's 15 October 2026 announcement described placing more links inside AI Overviews and AI Mode and showing them higher more often, so the goal is to be a source the answer links rather than a position in a list. That rests first on the page being crawlable and readable in the served HTML, because a page a crawler cannot fetch or parse is never a candidate to be linked, and only then on structured data and a clear answering passage.
Is appearing in an AI Overview the same as ranking first in Google?
No. The organic blue-link position and the link placed inside an AI answer are decided separately. A page can rank well and not be cited, or be cited without holding the top organic position, which is why the two are worth tracking as different outcomes rather than assuming one follows the other.
What are Preferred Sources in Google?
Preferred Sources is a reader-side feature that lets a person choose sites they want to see more often in Google's Search and AI experiences. Google said on 15 October 2026 that the number of people using it has grown by more than 7x since the start of the year, which is a count of people turning the feature on, not a measure of referral traffic to those sites.
Can I stop a page being used in AI Overviews?
Google's robots meta tag documentation states that nosnippet prevents the content being used as a direct input for AI Overviews and AI Mode, and that max-snippet limits how much may be used. Both require the crawler to reach the page first, so a page blocked in robots.txt is governed by the block, not by a snippet directive the crawler never receives.
See what AI can read on your site
Run a free scan and get a graded report of exactly what AI crawlers can and cannot read, with ranked fixes.
AI visibility
AI visibility is whether AI systems can access, read, and faithfully represent your web pages, the prerequisite for being cited in an AI answer.
Generative Engine Optimization (GEO)
Generative Engine Optimization (GEO) is the practice of structuring and publishing content so that generative AI engines can fetch it, read it in full, and cite it when they answer a question.
How to get content cited in Google AI Overviews: 0 of 81 home pages blocked their own snippet
Google publishes one hard eligibility rule for a supporting link in AI Overviews, which is that the page must be indexed and eligible to be shown in Search with a snippet. Lantad requested the home page of 120 real hostnames on 6 September 2026. Of the 81 that answered HTTP 200, none carried a nosnippet directive, three carried max-snippet and all three set it to -1, and 48 carried no robots meta tag at all.
Google shipped a preferred sources button, and a plain URL for sites that block third party scripts
Google Search Central updated its preferred sources documentation on 20 August 2026 to add an interactive button that loads a script from news.google.com, and named a plain URL alternative for sites that cannot run third party scripts.
Structured data for AI (schema markup)
Structured data is schema.org markup, usually written as JSON-LD, that labels a page's facts in a machine-readable format so AI systems can extract and attribute them without having to interpret the surrounding prose.
