Google spent October 15 telling the web it was adding more links to AI answers, and the same week it told Barry Schwartz it still has no click data to give you for them. Both halves of that are true, and taken together they are a clearer strategy brief than either one alone. The link surface in AI Mode got wider. The scoreboard did not arrive. Anyone planning AI Mode SEO around the first fact while ignoring the second is about to optimise for something they cannot see.
The announcement itself is real and reasonably generous. Nick Fox posted it on X on October 15, and the detail landed in a Google blog post under Hema Budaraju, VP of product management. Google described continued and substantial increases to the links embedded in words inside AI answers, said those inline links now appear higher in responses more frequently, gave the end of response Dive Deeper section image thumbnails and a more clickable design, and expanded its timely news carousels to more countries and languages.
Our read is that the headline change is not the one Google led with. Buried in the same post is a test of a new section at the top of an answer reserved for what Google calls the one or two official places on the web to complete a specific task or get the exact info you need, with context attached explaining why each link is useful. That is a two result slot at the top of an AI answer. Everything else in the announcement spreads attention thinner. This one concentrates it.
What Google changed in AI Mode and AI Overviews on October 15
Google announced five distinct changes to how links appear in AI Mode and AI Overviews on October 15, 2026, four of which make existing link formats more numerous or more visible, and one of which creates a new and much narrower slot at the top of an answer.
Separating them matters because they do not point the same direction. More inline links and a prettier Dive Deeper block widen the field. The official place test narrows it. The creator link test does something else again, which we will come back to, because it is the most revealing item on the list.
| CHANGE | STATUS | WHAT IT DOES TO YOUR ODDS |
|---|---|---|
| Continued and substantial increases to inline links inside answers | Shipping | More sources cited per answer, so a citation is easier to earn and worth less individually |
| Inline links placed higher in responses more frequently | Shipping | Position within an answer now matters, the way position within a result page always has |
| Dive Deeper section gets image thumbnails and a more clickable design | Shipping | A real click affordance at the end of an answer, though it sits after the user already has the answer |
| Timely news carousels expanding to more countries and languages | Shipping | Favours publishers users have set as Preferred Sources, which is a subscription style moat, not an SEO one |
| New top of answer section for the one or two official places to complete a task | Testing | A two result slot with explanatory context attached, which is the most valuable surface in the announcement |
| Creator and publisher names on video results become clickable | Testing | Points to a Google Search profile page first, not the creator's own site |
Google paired the announcement with two numbers about its publisher relationships rather than about traffic. It said usage of Preferred Sources has grown sevenfold this year, and that more than 400 publications worldwide now participate in its News AI pilot licensing programme, including Axios, CNN, the Financial Times and Forbes. Neither number tells a site owner anything about clicks, which is the pattern worth noticing early.
Why more links does not mean more clicks
A larger number of links inside an AI answer does not produce a proportionally larger number of visits, because the constraint on AI answer clicks is not how many links are available but how many users want to leave once the answer is already written for them.
The best public evidence on that constraint is Pew Research Center's browsing data study, summarised in detail by Sistrix, which tracked about 900 US adults across 68,879 unique Google searches. Roughly 18 percent of those searches produced an AI Overview. On searches without one, users clicked an organic result about 15 percent of the time. On searches with one, that halved to about 8 percent. Clicks on the links inside the overview itself ran at about 1 percent. Sessions ended outright on 26 percent of AI Overview searches against 16 percent of traditional ones. Google has publicly disputed the methodology, and that objection is worth stating alongside the numbers, but no competing dataset of comparable scope has been published.
Run the arithmetic honestly. If the click rate on in answer links is roughly fixed and low, then doubling the number of links in an answer does not double the traffic leaving that answer. It divides a largely fixed pool of departing users across more recipients. That is a redistribution. For a site that was never cited it is a genuine improvement, because some citation beats none. For a site that was already the single inline link on a query it is a dilution, and nobody announcing the change is going to describe it that way.
This is the same move we watched from the other direction when ChatGPT shifted its citations to the bottom of an answer: the number of sources an engine shows and the number of users it sends you are only loosely coupled, and engines adjust the first far more readily than the second. Treat link count as a visibility metric, never as a traffic forecast.
The official place box is the only slot that got more valuable
The new top of answer section, which Google is testing for the one or two official places on the web to complete a specific task or get the exact info you need, is the single element of the October 15 update that increases in value rather than diluting, because it caps the number of winners at two.
Everything about how it is described reads like a transactional intent slot rather than a research one. Complete a specific task. Get the exact info you need. Official. Google is also attaching context explaining why each link is useful, which means the slot carries an editorial endorsement that an inline citation does not. If you sell a service, run a tool, host a calculator, process an application or maintain the canonical document for a process, that is the surface to care about, and the ordinary inline link expansion is noise by comparison.
Ahrefs US volume and difficulty for the terms practitioners are actually searching around this shift, pulled October 17, 2026
Note the inversion in that chart, because it is the practical lesson. The broad awareness terms are the hardest to win and the commercial terms are the softest, which is the opposite of what most teams assume. Difficulty 13 on a term with 3,100 US searches a month and a 20 dollar average cost per click is not a defended position, it is an empty one. The same logic applies to the official place slot: the queries where Google wants to name an official destination are precisely the queries nobody optimised for, because they look too commercial to chase with content.
We argued a version of this in the reference pillar build for answer engine optimization, and the October 15 announcement sharpens it. An answer engine that nominates official destinations is going to nominate pages that look canonical: one page per task, a clear definition in the opening sentence, headings phrased as the questions users ask, and no gate between the user and the thing they came to do. That is a page structure decision, not a link building one.
What AI Mode SEO looks like after the link update
AI Mode SEO after October 15 has two jobs that pull in different directions: earning inline citations across many answers for coverage, and building a small number of unmistakably canonical task pages that qualify for the official place slot. Most programmes are only doing the first.
The split matters because the two jobs reward different work. Inline citation coverage is a volume game won by topical breadth, clean passage structure and being retrievable on many phrasings. The official place slot is a singularity game won by being the obvious destination for one specific job. A content calendar optimised purely for the first will never produce the second, which is how a site ends up widely cited and commercially invisible.
One more structural note from the announcement. Timely news carousels now favour Preferred Sources, the feature Google says has grown sevenfold in usage this year. That is a reader subscription mechanic, not an optimisation target. You cannot earn your way into someone's Preferred Sources list with better markup, which is worth saying plainly so nobody spends a quarter trying. Brand demand is the input there, and it belongs in a brand plan rather than a technical backlog.
How to measure AI Mode SEO without click data
Measuring AI Mode performance currently requires substituting first party and observational sources for the click data Google does not provide, because Search Console reports AI surface activity inside its totals without letting you isolate clicks, impressions or click through rate for AI Mode.
Schwartz put the complaint precisely in his write-up of the announcement, and it deserves repeating: Google keeps making these links more prominent, and clicks remain missing from the reports, even though clicks are the metric publishers most want and the headline metric in Google Ads. Google's response, given to Search Engine Roundtable, was that it is working with site owners to decide which insights would be most useful and will add more metrics over time. That is not a commitment with a date on it. Plan as though the gap is permanent and treat any future reporting as upside.
| WHAT YOU WANT TO KNOW | WHAT YOU CAN ACTUALLY OBSERVE | HOW MUCH TO TRUST IT |
|---|---|---|
| Did AI Mode send me traffic | Total Search Console clicks including AI surfaces, not separable | Directional only, never attribute a delta to AI Mode from this alone |
| Am I being cited in AI answers | Manual and automated prompt sampling against your target query set | Reliable for presence, noisy for share, so sample consistently and track trend not snapshot |
| Is retrieval happening at all | AI crawler hits in server and CDN logs, by directory | High. Your own infrastructure, unaffected by anyone else's reporting choices |
| Which pages the engines prefer | Which URLs appear as the cited link, logged over time | High, and more actionable than any aggregate score |
| Am I in the official place slot | Prompt sampling on task phrasings, checked for the top of answer box | Reliable once the test is widely live, worth baselining before it is |
| Did the October 15 change help me | Nothing in Google's reporting will answer this | Zero. Do not build a slide claiming it did either way |
This is the second measurement surface to degrade or stall in two weeks, which is starting to look like a trend rather than a run of bad luck. We covered what the generative AI report in Search Console actually measures in September, and the rank tracking data problem earlier this week. The consistent lesson across all three is that instruments you do not own get less reliable at exactly the moment the surfaces they measure get more important. Teams we run reporting and analytics work for have mostly made the switch to log based retrieval reporting already, and the ones that have are the only ones who can say anything factual about October.
Set the baseline now, before the official place test goes wide. Once that box is live across your category you will want to know whether you held the slot from the start or lost it, and there will be no retrospective data to recover.
What to do about it this week
The correct response to the October 15 update is a short, concrete sequence: identify your task pages, make one of them unmistakably canonical, baseline citation presence while the new slot is still in testing, and stop reporting AI performance through metrics Google has declined to provide.
None of this requires waiting for the test to finish rolling out. The page structure that qualifies for an official place nomination is the same structure that reads well as an inline citation, so the work pays off either way. That is unusually convenient and worth moving on while competitors read the announcement as a story about inline link counts.
The broader point, and the reason we think this update rewards attention rather than a shrug: Google has now expanded the link surface twice this year while holding the click reporting flat, and on October 16 Search Engine Roundtable noted a new Gemini model landing in AI Mode on top of it all. The surface keeps moving and the instrument panel keeps not arriving. Programmes built on earning citations and measuring them yourself absorb that fine. Programmes built on waiting for Google to report AI clicks have now been waiting a year. Serious generative engine optimization services work the first way, which is less a methodology preference than an acknowledgement of who controls the data.
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Josh leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.