September 25, 2026: Google’s AI Overviews are displaying more “links” — and some of those clicks don’t go to your site. They open Google’s own AI Mode experience instead. That changes what “winning a link” actually means, and it changes how you should read Search Console this month.
The practical takeaway is blunt: if your team is celebrating AI Overview visibility without confirming outbound clicks, you may be celebrating Google engagement, not business outcomes.
What changed (and what Google is counting)
Search Engine Journal reports that AI Overviews can include link interactions that either: (1) send a user to a publisher’s page, or (2) route the user into Google’s AI Mode. Both look like link opportunities on the surface, but they behave differently for traffic.
Google also measures these interactions differently in Search Console. So two sites can both “appear” in AI Overviews, but one gets actual visits and the other mostly triggers on-SERP exploration.
This isn’t a minor reporting quirk. It’s a product direction: Google is expanding in-SERP journeys. Your SEO plan needs to treat “AI Overview link” as two separate outcomes: outbound referral vs Google-contained session.
Why business owners should care (not just SEOs)
If you sell anything that requires consideration — B2B services, higher-ticket consumer services, anything with a sales call — your funnel doesn’t start when Google re-explains your page. It starts when someone hits your site, sees proof, and converts.
AI Overviews that route into AI Mode can still have brand value. But brand value is not pipeline. If your reporting doesn’t separate outbound clicks from on-SERP engagement, you’ll mis-allocate budget: doubling down on content that “shows up” while your leads flatten.
This is especially risky right now because many teams are also dealing with ranking turbulence around spam enforcement. When traffic moves, attribution gets messy, and it becomes easy to blame the wrong thing. If you’re already monitoring volatility due to the September 25, 2026 search forum recap covering the spam update rollout window, you need clean instrumentation so you can tell “we lost clicks” from “Google kept the clicks.”
Our read: who wins, who loses
Winners: sites that answer one question better than anyone else, with a crisp snippet-friendly structure (definitions, steps, comparisons) and strong entity signals (clear product/service pages, consistent naming, structured data where it actually matches the content). They’ll be pulled in as citations and still earn the outbound visit when the user needs depth.
Also winners: brands with obvious trust markers on-page — real author bios, policies, pricing ranges, location specifics, and first-party evidence. AI Overviews can summarize claims, but users still click when they need to verify.
Losers: thin “SEO middleman” pages whose only job was to intercept long-tail queries and repackage what’s already known. AI Mode is basically built to replace that layer.
Everyone will miss this: the metric you should chase is not “AI Overview impressions.” It’s “AI Overview impressions that generate outbound clicks and assisted conversions.” If you can’t report that, you’ll optimize the wrong pages for the next 90 days and call it progress.
What this means for your site
- Split your reporting: create two dashboards: (1) pages getting AI Overview exposure, (2) pages getting AI Overview exposure plus measurable outbound clicks and engaged sessions (GA4).
- Rewrite intros for citation extraction: your first 80–120 words should answer the query directly, then expand. We’re updating client pages to lead with a one-sentence answer, not a brand preamble.
- Strengthen “click-for-proof” sections: add the thing AI can’t credibly compress: screenshots, methodology, pricing examples, code snippets, calculators, before/after photos, downloadable templates.
- Audit structured data for match quality: keep only schema that exactly reflects the page. Misaligned markup tends to backfire over time, especially as Google tightens interpretation.
- Track query classes, not just pages: group queries into “quick answer” vs “needs depth.” Expect quick-answer queries to drift toward AI Mode; fight for depth queries where a click still matters.
What we’re changing on client sites this week (effort vs impact)
This is not a “rebuild the whole site” moment. It’s a “tighten the parts Google extracts and the parts humans verify” moment.
1) We’re rebuilding page headers for extractability (1–2 days per priority page). For service pages, we move the core definition and differentiator above the fold. For product-led pages, we add a simple “What it is / Who it’s for / What it costs” block early. If you’re planning a rebuild anyway, do it with a performance baseline; teams doing London-focused web builds often forget that copy structure changes can move the needle faster than component changes.
2) We’re adding proof assets that force the click (2–5 days per cluster). Think: interactive tools, comparison tables that reflect your real options, short video demos hosted on your site, and transparent constraints (“this approach won’t work if…”). AI Overviews can summarize benefits; they struggle to replace evidence.
3) We’re tightening technical hygiene on money pages (half-day to 2 days). Canonicals, internal linking to the commercial page, and cleaning index bloat. If your dev work is queued, it’s worth aligning it with SEO priorities; we see this most often when teams reference our US software build pricing guide to budget a feature, but don’t budget the 1–2 days to ship the SEO plumbing that lets it rank.
4) We’re separating “visibility” from “visits” in stakeholder reporting (same day). This is where most teams get burned. A chart showing AI Overview impressions will go up. Leads may not. Your CEO will notice in 6–10 weeks.
One simple table to stop bad decisions
| Signal | What it might mean now | What to do |
|---|---|---|
| AI Overview visibility up | You’re being cited, but users may stay in Google | Verify outbound clicks + assisted conversions before scaling content |
| Clicks flat, impressions up | More on-SERP engagement / AI Mode routing | Add proof assets and stronger commercial intent pathways |
| Clicks down after a rewrite | You improved extractability but weakened the reason to click | Re-introduce differentiation, examples, and “why us” above the fold |
If you do nothing for 90 days
You’ll likely report “SEO is stable” because impressions and appearances look fine, while inbound leads quietly soften. By the time it’s obvious, competitors will have rebuilt their top clusters around proof and intent, and you’ll be stuck reworking content under pressure.
There’s also a compounding effect: if users increasingly satisfy curiosity inside AI Mode, the remaining clicks skew toward higher intent. That’s good news — but only if your landing pages are built to convert that intent. If your pages are generic, you’ll lose the best traffic you still get.
If you operate across markets, align the playbook. We’re applying the same measurement split for teams running US pipeline-focused SEO programs, and for EU clients working with our team on SEO work in German-language markets. The mechanics differ by SERP, but the principle is identical: measure outbound clicks, then optimize for the queries that still produce them.
Next step: pick 10 pages you believe “win” AI Overviews, and audit them for (1) extractable answers, (2) irresistible proof, and (3) clean measurement. If you want us to sanity-check the list and prioritization, send the URLs and your last 28 days of Search Console + GA4 exports.
Primary source: Search Engine Journal’s report on AI Overview links routing to AI Mode vs the web. Related Google context on Search Console’s broader reporting direction: Google Search Central on consolidating performance across social and video platforms.
Primary source: Search Engine Journal
