The number that should scare you
On informational queries where Google surfaces an AI Overview, the average click-through rate to the top organic result has dropped from roughly 28% in 2024 to roughly 9% in 2026 across the verticals we measure. That’s not a slow erosion — that’s a cliff.
The buyer reads the AI box, gets their answer, and closes the tab. The blue links underneath barely register. Some of them never load — the cursor moves to the address bar before the page paints.
If you’re still publishing content optimised for “rank #1 on Google” without checking whether AIO already eats the answer, you’re funding traffic that no longer arrives. The pages still get crawled, indexed and ranked. They just don’t get clicks.
Which page types lose the most
The pattern is consistent across the niches we track. AIO eats clicks on anything Google can answer in sixty words.
Definitions — “what is X”, “X meaning”, “X explained”. AIO returns a tight 40-word answer with a Wikipedia-style framing. Click-through to your 1,800-word “ultimate guide to X” collapses. The user got what they needed. There is no second question.
Simple how-tos with three or four steps — “how to reset a router”, “how to format a USB drive on Mac”. AIO lists the steps inline. The “click” used to come from users wanting to verify against a screenshot — but AIO now embeds inline images for the most-searched of these.
Statistics roundups — “average customer acquisition cost SaaS 2026”, “how many people use ChatGPT”. AIO pulls the number, names the source, and the user is done. Your 2,400-word “state of X” report with thirty stat citations gets paraphrased into one box.
Comparison-of-features matrices when the answer is one row — when the user asks “does Notion support offline mode”, AIO returns yes/no with one sentence. The comparison post that took two days to write doesn’t get the click.
Date and time-sensitive trivia — opening hours, public holiday lists, currency symbols, ISO codes. These were never high-value clicks, but they were a chunk of impression volume that’s now gone.
If your editorial calendar still leans on those page types as primary traffic drivers — that’s the bleeding wound. They still rank. They still impress. They no longer convert into sessions.
Which page types still earn the click
The mirror image of the above. AIO can’t replace pages where the user needs to land somewhere to complete the task.
| Page type | Why AIO can’t kill it |
|---|---|
| Transactional (pricing, signup, checkout) | The action requires the destination |
| Comparison “X vs Y” with opinion | AIO refuses to take sides on competing products |
| Deep technical with code, config, schemas | AIO truncates code blocks; users need to copy-paste |
| YMYL with a named source (legal, medical, financial) | Users want to verify the human author behind the claim |
| Editorial / opinion (reviews, takes, predictions) | AIO doesn’t synthesise opinion well — it hedges |
| Live data (prices, inventory, availability, schedules) | AIO can’t reliably pull real-time data into the box |
| Tools and calculators | The user needs to input numbers, not read about them |
| Original research with proprietary data | AIO cites you instead of replacing you — and the cite is the click |
That’s where your editorial budget should sit in 2026. Not on definitions. Not on “what is X” listicles. Not on stat roundups someone else will reformat in three weeks anyway.
Measuring erosion in Search Console
Google won’t hand you a clean “AIO impression” column. You have to triangulate.
The impression-to-click divergence test. Pull a 90-day window in GSC. Group queries by intent type — informational vs transactional. Plot impressions and clicks on the same chart. If informational impressions are flat or growing while informational clicks are falling — that’s AIO erosion, not a ranking drop.
The position-stability filter. Filter queries where average position is stable (delta within ±0.5) but CTR has fallen by 30%+ over 90 days. Position didn’t change. Something else absorbed the click. That something is almost always AIO on informational queries.
The page-level cohort. Pick ten pages by template — five definition pages, five transactional pages. Compare their CTR trend month over month. If definitions are falling 4–8 points while transactional is steady — confirmed pattern.
The AIO-presence indicator. Manual or via a tool like Searchable, Profound or Orion. Every Friday, fetch the SERP for your top-30 tracked queries, log whether AIO appears, log whether you’re cited. Match against GSC clicks. The correlation between “AIO present” and “CTR collapsed” will be obvious in three weeks.
GSC alone is not enough. The Performance report doesn’t tell you why CTR dropped — it just shows that it did. The AIO-presence layer is what makes the diagnosis. We covered the broader stack in how to measure AI citations.
The brand-search lift nobody talks about
Here’s the part that gets missed when teams panic about CTR loss. When AIO names you in the answer — “according to Answerly”, “per Gofaizen & Sherle”, “as reported by [your brand]” — branded search lifts. Sometimes a lot.
Across the engagements we run, brands that earn AIO citations on their priority queries see branded-search volume rise +12% to +35% over the following 60 days. The user reads the answer, sees your name, doesn’t click in that session, and Googles your brand directly two days later when the buying decision moves up the priority list.
The click you lost on the informational query came back as a branded session — which converts at 4–10× the rate of cold informational traffic anyway.
This is why “AIO killed our traffic” is the wrong framing if your brand is being cited. The traffic shape changed. The funnel shape changed. Total qualified traffic often went up. You just have to measure it across two funnels instead of one.
To track it: pull branded-search volume in GSC (queries containing your brand name), pull AIO-citation count from your measurement tool, plot them on the same axis with a 30-day lag. The lift shows up. If it doesn’t, your AIO citations aren’t sticking — and that’s a separate problem to fix.
Tactics that actually work in 2026
Not “how to defeat AIO”. Nobody defeats AIO. The tactics are about being the cited source on informational queries while keeping the click flowing on transactional ones.
Be the cited source, not the destination, on definitions. Stop trying to rank a 2,000-word “what is X” page hoping for clicks. Optimise it for AIO citation — Quick Facts table, X-is-Y intro, FAQ-style direct answers. Concede the click on that query. Earn the brand mention. Convert via branded search later.
Structure for partial extraction on the pages you do want clicks on. AIO pulls a paragraph. Make sure the paragraph it pulls is incomplete on its own — the second half of the answer requires the click. “There are seven steps. The first three: …” — AIO will quote the three. The user clicks for the other four.
Price, availability and live-data content AIO can’t fully answer. Pricing pages with tables. Inventory status. Booking calendars. Stock checkers. AIO will summarise that “prices range from $X to $Y” but won’t give the actual number for a specific config. The click is forced.
Comparison content with strong opinion. “X vs Y: which to pick” with a named author, an actual recommendation, and a decision matrix. AIO hedges on these. The user wants the take. The take requires the click.
Original data with proprietary numbers. If you ran a survey, did a study, scraped your own dataset — AIO cites you, the user clicks for the methodology, and you get the click and the citation.
Schema for the right things. FAQPage and HowTo schema increase your odds of being the cited source. Article schema with named author increases your odds of being chosen on YMYL. We unpacked this in the Google AI Overviews 2026 piece.
When to give up on a page type entirely
This is the contrarian part most agencies won’t tell you.
If you have a category of pages — informational long-tail definitions, “what is X” articles, basic how-tos — that fully meet all three of these criteria, stop producing them:
- AIO appears on 70%+ of the queries
- CTR has fallen below 5% over 90 days
- The page type has zero observable conversion to revenue or qualified lead
There is no version of “more content” that fixes that. The content economics are gone. The traffic that used to subsidise the production cost is now subsidising Google’s answer box. Reallocate the budget to the page types that still earn the click — comparison, transactional, original research, deep technical.
Most agencies won’t recommend this because their retainer is sized around content volume. Cutting production cuts billable hours. We recommend it because we measure outcomes, not output. The Growth tier already builds in fewer-but-better page production for exactly this reason.
The two-funnel reframe
Here is the take. If you treat AI Overview as Google’s algorithm change to fight, you lose. You spend the next two years writing more, longer, fresher informational content that AIO digests and serves without you.
Treat it instead as two separate funnels.
Funnel one — citation channel. AIO is a citation surface, like Wikipedia or a top-tier publication. You optimise for being quoted. The KPI is citation count and share-of-voice in AIO answers, not click-through. The downstream effect is brand awareness and branded-search lift.
Funnel two — conversion channel. Transactional, comparison, deep technical, YMYL, original data. The KPI is clicks, sessions, conversions. AIO doesn’t compete here in any meaningful way — and where it does, the click is forced because the user can’t complete the task in the answer box.
The brands winning in 2026 separate those two funnels deliberately. They write differently for each. They measure differently for each. They allocate budget differently for each.
The brands losing in 2026 still treat the SERP as one funnel — and they’re being slowly digested.
What to do this month
One week of measurement, three weeks of action.
- Pull 90 days of GSC data, segment by intent type, identify the five page templates with the sharpest CTR collapse
- Set up an AIO-presence indicator (manual top-30 SERP scan every Friday) and confirm the correlation
- Pick three page types to concede the click on — restructure them for AIO citation, accept the funnel-one outcome
- Pick three page types to defend — add price/comparison/opinion/data elements that force the click
- Track branded-search volume on a 60-day lag — if AIO citation is rising and brand search isn’t, your citations aren’t sticking and that’s the next thing to fix
By month two, you’ll know which funnel is doing what. By month four, the budget reallocation pays for itself in qualified sessions. We’ve watched this play out across the cases on our portfolio — the pattern is consistent.
The CTR cliff is real. The reframe is the work.