Liz Reid Just Conceded AI Overviews Skip the Browsy Half of the SERP
Google's Liz Reid told the Search Off the Record podcast that AI Overviews still leave "browsy" queries to the full SERP because users want to compare options rather than accept one synthesized answer. Seer Interactive's 2026 audit found AI Overviews trigger on 36% of informational queries, 8% of commercial, and 5% of transactional. The queries Google is actually eating are the specific informational ones, not the shoppy ones, and that flips most SEO triage I've seen this quarter.
What Liz Reid actually said about browsy queries
Reid is VP of Search and the public face of how Google is reasoning about AI Overviews and AI Mode. On the latest Search Off the Record episode, she described "browsy" queries as the under-specified, exploratory ones where the user genuinely wants to scan multiple options. Discovery intent. The classic example: someone typing "best running shoes" who doesn't actually want one synthesized answer, they want to look at five.
Her framing, as summarized by Search Engine Journal, is that those queries still favor the full SERP because the answer the user wants is "show me a set." AI handling, by contrast, gets routed to queries where the user has done the work of being specific. Long-tail. Multi-clause. The kind of thing you'd type if you already know what you're trying to figure out.
That's a useful framing if you want to feel reassured. It also quietly redraws which third of your organic traffic is in the danger zone. Reid did not put it that way, of course. Google's official AI-in-Search positioning is that everything works in concert. The thing she conceded is that the routing rules are visible enough now to plan around.
The intent split is doing more work than Google's framing admits
The Seer audit numbers are the part worth printing out. AI Overviews trigger on 36% of informational queries, 8% of commercial-investigation queries, and 5% of transactional queries. Queries 8 words or longer are roughly 7x more likely to trigger an AIO than 1-3 word queries. Search Engine Land's broader appearance data puts the overall AIO rate around 13% of searches, which lines up with the weighted average across intent types.
What that actually means: the panicking I keep seeing about AI Overviews wiping out organic clicks is partly miscalibrated. If your money pages are commercial or transactional queries (someone searching "buy X," "X pricing," "X vs Y for purchase"), you're sitting in the 5-8% trigger zone. Not zero, but nowhere near the apocalypse number that gets circulated.
The pages quietly bleeding are the ones nobody wants to admit are pages: the long-tail informational guides. The "how does X work" content. The 1,800-word explainers that ranked because nobody else wrote them. From what I've seen across a few SEO audits this spring, that's exactly the bucket where the click loss is concentrated, and it's also the bucket most teams underweight when they triage.
So the question isn't "is AIO eating my traffic." It's "which third of my traffic was always going to be eaten first."
Where this actually leaves your money pages
Look at the routing logic Reid is hinting at and the picture clears up. Discovery-mode browsy queries: full SERP, AIO mostly absent, comparison content still wins. Specific informational queries: AIO eats the click. Commercial and transactional intent: SERPs largely intact because Google still wants users to click into shopping comparisons and product pages.
The implication is that the SEO playbook for the next two quarters splits cleanly into two buckets. Bucket one is your commercial and transactional content. That stuff is mostly fine. Keep the conversion pages tight, keep the buyer-intent landing pages updated, but you don't need to rebuild the funnel. Bucket two is your informational top-of-funnel content. That's where the work is.
And the work isn't to write more or to bulk up. It's to figure out which informational queries you're winning are browsy (and therefore safe) versus specific (and therefore in the eat zone). A query like "marketing automation tools" is browsy. A query like "how does HubSpot's lead scoring algorithm calculate a contact's grade" is specific. Same topic, different intent. Different AIO risk. The reframing move is to keep producing exploratory comparison content that wants to be a SERP, not a single answer.
This is also the part where the earlier Google concession that AI Overviews and AI Mode run on separate stacks matters. Different routing logic. Different ranking signals. Reid's intent split applies to AIO. AI Mode is its own animal.
The 30-minute audit that maps your exposure
Pull last 90 days of Search Console data. Sort by clicks. Take the top 50 queries driving organic traffic. Bucket each one as informational, commercial, or transactional based on what someone typing it actually wants. Then add a fourth column: query word count.
The exposure threshold I'd watch is whether more than 40% of your top-50 traffic comes from informational queries that are 8 or more words long. If yes, you have meaningful AIO exposure and the click decay is probably already happening in your data, just spread thin enough to look like noise. If under 40%, you're in better shape than the average panic post implies.
For the high-exposure queries, sample 10 of them in an incognito Google search and note which ones trigger an AIO box today. Coverage from Search Engine Land shows AIO appearance is uneven by category, so manual sampling beats assumed coverage. The ones that do trigger are your decay candidates. The ones that don't, even if they're 8+ words, are still browsy enough to escape, and you've identified the survivors.
The pattern to look for in the data is impressions holding flat while clicks slide on the long informational queries. Same pages, same rankings, fewer clicks. The Seer audit calls that the AIO signature, and it's the cleanest signal because rankings staying put rules out the usual core-update suspects. If your GSC export looks like that across a dozen of your top informational queries, the audit isn't theoretical, it's already overdue.
The action this week: get the spreadsheet, do the bucketing, mark the decay candidates. Don't rewrite anything yet. The audit is the deliverable. The rewrite plan comes after you actually know which 12 articles are in the eat zone, not all 200.
Where Reid's framing still hides the trade
The piece of this Reid did not address, and the part I keep coming back to, is that "browsy" is a moving target. Google's models decide what counts as browsy. That decision can shift quietly between core updates. A query that gets a full SERP today might get an AIO tomorrow if Google reclassifies the intent, and the only way you'll find out is by watching impressions hold while clicks drop. Which is exactly the pattern the brand-search-vs-backlinks data was already showing for AI citations.
The audit above is still worth doing. It tells you where you stand right now. The thing it can't tell you is when the line moves. From what I've seen, the only real defense is to keep producing comparison-shaped content that's structurally hard to synthesize into a single answer, even on informational topics. Lists. Pro/con tables. "Five tools and the one trade-off each one makes." Content shaped like a SERP, not like a paragraph.
Reid drew a line. Worth knowing it exists. Just don't assume the chalk holds.
Notice Me Senpai Editorial
