Google's AI answers now link your brand name higher. Which of your pages does the link open?

KEY TAKEAWAYS

1. Google's 15 October change helps most on task queries with an official answer, not on discovery prompts.
2. The measurable problem is destination. Inline links on brand names often open a homepage or the wrong page.
3. Give each customer task one owner page and audit which URL each AI answer opens.
4. Search Console cannot show the effect yet, so run the audit yourself and log results from 15 October.

On 15 October 2026, Google said AI Mode and AI Overviews will show inline links higher in answers and more often. It also began testing a section at the top of some answers that lists official pages for a task. Most coverage has treated this as more traffic for everyone. The evidence points somewhere narrower. The change matters for one kind of question, and the first test shows the links often open the wrong page.

What did Google actually change?

Hema Budaraju, a vice president of product management, announced five changes on Google's blog:

1. Inline links. Google reported "continued and substantial increases" in links placed inside answer text, now shown higher and more often.
2. Official pages section. A test. For tasks where only one or two official sources exist, some answers show those pages at the top, each with a note on why it helps. Google's example image, shown by Search Engine Journal, lists airline refund links under the label "Quick results from the web."
3. News carousels. Expanding to more locations and languages, and featuring a reader's Preferred Sources. Google said Preferred Sources use has grown more than 7x since the start of 2026.
4. Related article links. Thumbnails and a more clickable design at the end of some answers.
5. Creator names. A test of clickable creator and publisher names on AI Mode video results, leading to Google Search profiles.

Google gave no figures for the size of the inline link increase, no markets, no dates and no criteria for what counts as official. The direction is not new. Google announced more inline links and hover previews on 6 May. Peec AI data reported by ALM Corp showed the share of AI Overviews with an external link inside the answer text sat near zero until about 11 September, then rose to 26.2% on 20 September. Peec did not publish its sample, so read that as a trend.

Which questions does this change reach?

Not all of them. The official pages section, by Google's own description, is for tasks with one or two official answers. Think refund policy, cancellation, login, pricing, API documentation, booking or a returns form. For those, the brand is the source by definition.

Discovery questions behave differently. Shero Commerce studied purchasing questions across 60 product categories in Google AI Mode, ChatGPT and Perplexity, as reported by Search Engine Journal. Only 2.8% of 1,851 cited sources were brand-owned pages, and 59% were third-party sites. A separate preprint from Ranqo, summarised by Lisa Peyton, covered 102 brands on ChatGPT, Gemini, Perplexity, Claude and Grok. It found the tracked brand's own site was 2.9% of citations. Surfer's September data on 1.2 million AI Mode answers found google.com alone took 13.82% of all citations.

So the October changes help most where you already own the answer. For "best payroll software", a higher inline link mostly sends the reader to whichever review site or listicle holds the claim.

Do higher links bring clicks back?

Nobody can say yet. Google published no click data with the announcement, and Search Console's generative AI report still has none. The best evidence on clicks comes from two randomized studies, both run before October.

• Agarwal and Sen (Indian School of Business and Carnegie Mellon) randomized 1,065 US desktop Chrome users from 7 January to 10 February 2026. When an AI Overview appeared, organic clicks per search were 0.37, against 0.62 when it was hidden. The authors report a 39.8% drop in organic clicks.
• Wang and colleagues (Penn and Northeastern) ran a preregistered seven-day experiment in March 2026 with 1,100 active participants. Sending all searches to AI Mode cut external click-through by 18.8 percentage points.

One detail matters for this article. According to PPC Land's reading of the June revision of Agarwal and Sen, the effect came from informational queries, 71% of the sample. Navigational and transactional queries showed effects that could not be distinguished from zero. Those estimates are noisy, and the earlier seminar version said the breakdown was still to come. Still, the pattern fits: when a person needs to do something on a specific site, the answer does not replace the visit.

Google's documentation says clicks from results with AI Overviews tend to be higher quality. Agarwal and Sen found no difference in bounce rate, back-button returns or time on page. The only randomized data on engagement does not support Google's claim.

Where do the inline links actually go?

This is where the risk sits. The day after the announcement, Flying Cat ran seven B2B software task prompts in AI Mode and published the results in The Search Spot:

• The official pages section appeared in 0 of 7 answers.
• The answers carried 27 inline links. 13 pointed to the product company's own site.
• 9 of the 27 opened a homepage. In three answers, the help page that explained the task appeared only as a source chip.
• Some links missed. An "application" link opened a careers page, and a login link opened an analytics subdomain.
• The free-trial prompt did best. Three of four links opened trial, demo or access pages.

By our calculation, that is about 48% of links to the brand's own site and 33% to a homepage. It is one run, one signed-in US account and seven prompts, so the numbers carry wide error. The shape is still useful. Inline links attach to brand names, and a brand name tends to resolve to a homepage unless one page is the obvious answer for the task.

This also reconciles the two sets of numbers. Owned sites are about 3% of citations on discovery prompts and close to half of inline links on task prompts in one small test. Both can be true because the questions differ.

A worked example: scoring link destinations

The numbers below are hypothetical. They show the method, not a benchmark.

A software company lists five customer tasks and runs each prompt four times in AI Mode. It records every inline link that names the brand and the URL it opens.

The task match rate is 11 of 24, about 46% (our calculation). The weak rows tell you where to work. Cancellation and export sit under general help hubs, so the homepage wins. The API row splits between two documentation versions, which is a duplicate URL problem. Pricing and trial already work because each has one clear page.

Weight the fixes by what a wrong landing costs. A cancellation link that lands on the homepage creates a support ticket. A pricing link that lands on the wrong page can lose a sale.

What should you do for each kind of query?

Google's documentation says there are no extra requirements for AI features beyond being indexed and eligible for a snippet. It asks that structured data match visible text and that Business Profile details stay current. In digital authority terms, task pages sit with search presence and brand entity. Discovery answers are still a press coverage problem, and branded AI Overviews draw on third-party sources too.

Why can't Search Console show this yet?

Three reasons. The generative AI report shows impressions but no clicks. Every link inside an AI Overview takes the position of the whole block, as John Mueller confirmed in September, so a link moving higher will not change reported position. And a follow-up asked in AI Mode counts as a separate query. That is why an AI impression is hard to value. For now, a self-run audit is the only direct read on where links go.

A seven-step task page audit

1. List 20 customer tasks. Pull them from support tickets, sales calls and site search. Use the customer's words.
2. Name one owner page per task. If two pages compete, merge them or redirect one.
3. Make the task obvious. Put the task in the title and first heading, and the steps or answer near the top.
4. Run each task prompt in AI Mode and AI Overviews. Use a signed-out browser, two or three runs per prompt, and note the market.
5. Log every inline link on your brand name. Record the URL it opens: task page, homepage or wrong page.
6. Fix by cost. Start with tasks where a wrong landing costs a sale or a support ticket.
7. Annotate 15 October in analytics and re-run monthly. Track sessions to each owner page, and watch for the official pages section as it rolls out.

If you want this audit run alongside a digital authority score across ChatGPT, Perplexity and Google's AI surfaces, join the waitlist.

Questions

What is the official pages section in Google AI Mode?

A test Google announced on 15 October 2026. For tasks where only one or two official sources exist, some answers show those pages in a section at the top, with a note on why each helps. Google has not said how it picks official pages, which markets get it, or when.

Do more inline links mean more clicks from AI answers?

Not proven. Google published no click data, and Search Console still shows no AI clicks. Two randomized studies from early 2026 found AI answers reduce outbound clicks overall. Neither tested the October link changes, so whether higher links recover clicks is unmeasured.

Why does Google's inline link on our brand name open the homepage?

Google has not explained how it picks a destination. In one small October test, 9 of 27 inline links opened a homepage, and some help pages appeared only as source chips. A single clear page per task, with the task in its title, gives the system an obvious target.

Will the official pages section help with best of or comparison prompts?

Unlikely. Google describes it for tasks with one or two official sources. On purchase discovery prompts, brand-owned pages are only about 3% of citations in two 2026 studies. Those answers are still won through third-party coverage.

Sources

1. Google, More ways to connect with the web from Search blog.google
2. Search Engine Land, Google updates AI Mode and AI Overview with more links, context, carousels and information on creators searchengineland.com
3. Search Engine Journal, Google puts inline links higher in more AI search responses searchenginejournal.com
4. The Search Spot, Google adds more links to AI Mode and AI Overviews (October 15, 2026) thesearchspot.substack.com
5. PPC Land, Google pushes links higher in AI answers as Preferred Sources use rises 7x ppc.land
6. BetaNews, Google moves inline links higher in AI Mode and AI Overviews betanews.com
7. Search Engine Land, Google updates links within AI Overviews and AI Mode searchengineland.com
8. ALM Corp, Some links in Google AI Overviews open AI Mode instead almcorp.com
9. Agarwal and Sen, The impact of Google AI Overviews on publisher traffic and user experience. Evidence from a field experiment (seminar version) tse-fr.eu
10. PPC Land, Researchers find Google AI Overviews cut publisher clicks 39.8% ppc.land
11. ISB Institute of Data Science, AI summaries in search and their impact on publisher traffic isb.edu
12. Wang, Gleason, Bart, Wilson and Metaxa, AI in search reduces publisher referrals without improving user experience (arXiv 2608.18352) alphaxiv.org
13. Relevant Audience, AI Mode field experiment. Clicks drop 18.8 points relevantaudience.com
14. Google Search Central, AI features and your website developers.google.com
15. Search Engine Journal, AI tools recommend brands but cite other sites, data shows searchenginejournal.com
16. Surfer, The 50 most-cited domains in Google AI Mode (September 2026) surferseo.com
17. Lisa Peyton, AI cites company websites. Just not always yours lisapeyton.com
18. TechWyse, Google's John Mueller confirms Search Console AI reporting cannot track position within AI Overviews techwyse.com

NEXT STEP

See where your brand stands.

Your Digital Authority Score, who AI recommends for your target questions, and the placements that close each gap. Your first report is free.

Get your free report →