What AI Overviews Do to Your Click-Through Rate

Info Edge reported that billings at Shiksha, its education vertical, fell about 22.8 percent year on year to roughly 34.6 crore rupees in Q1 FY27, down from 44.8 crore. Brokerages tied the fall to AI driven search cutting referral traffic. Every other Info Edge segment grew that quarter. That is the clearest Indian number available on what AI search does to a business built on organic discovery.

Most brand teams are seeing a softer version of the same thing. Rankings look fine. Traffic does not. The confusion comes from watching the wrong number.

The damage does not look like a ranking drop

An AI Overview does not remove you from the results page. It sits above it. You still rank third. You still get counted as an impression. The user reads the summary and leaves.

So the signature is specific and easy to miss:

• Impressions flat or rising
• Clicks falling
• Average position stable or slightly improved
• The loss concentrated on question shaped queries, not on brand or product queries

If you look only at sessions in analytics, you see a decline and assume you were outranked. You then spend a quarter rebuilding content that was never the problem.

Which query types actually lose

The loss is not evenly spread. Independent studies disagree on magnitude and agree on shape.

Pew Research Center tracked around 68,000 real searches and found users clicked a result 8 percent of the time when an AI summary appeared, against 15 percent when it did not. Seer Interactive, working across a much larger impression set, measured organic click-through on informational queries falling by roughly 61 percent with an AI Overview present. Amsive, across about 700,000 keywords, put the blended decline nearer 15 percent. Google has publicly questioned some of this methodology.

Treat the range as the honest answer. Nobody has a clean number, and yours will differ by category.

The pattern underneath is consistent:

• Informational queries lose the most. A question like what is whey protein isolate is exactly what a summary closes.
• Transactional queries lose least. AI Overviews trigger far less often on buying intent, and a summary cannot complete a purchase.
• Branded queries can gain. Amsive found branded searches with an AI Overview showing higher click-through, because the panel works as confirmation that you are real.

Isolating it in Search Console

Do this before you change anything. Search Console is free and it is the only place you can see impressions and clicks together for the same query.

• Open Performance, Search results. Compare the last 90 days against the same 90 days last year to strip out seasonality.
• Turn on all four metrics: clicks, impressions, click-through rate, average position.
• Export to Sheets. Compute the change in clicks and the change in impressions per query, side by side.
• Sort for queries where impressions held within about 10 percent but clicks fell more than 25 percent. That is your suspect list.
• Tag each suspect query as informational, commercial, transactional or branded. The concentration becomes obvious inside a hundred rows.

Then read the newer report. Google launched a generative AI performance view in Search Console on 3 June 2026, with data starting 18 May 2026 and no historical backfill. It shows impressions from AI Overviews and AI Mode. It does not yet break out clicks. So it confirms exposure, not outcome. Read it as a presence check.

One more pass worth thirty minutes: take your top twenty commercial queries and run them manually on mobile, in India, logged out. Note whether an AI Overview fires, whether you are cited inside it, and how far down the first organic link sits. That tells you more than most tools currently do.

Which pages to defend first

You cannot defend everything. Sort by what the loss actually costs.

• Pages that still convert. A category page losing 15 percent of clicks matters more than a glossary post losing 60 percent.
• Pages one step from a purchase. Comparison pages, best-for pages, pricing pages, size and compatibility pages. Buyers still click these because they need to verify something specific.
• Pages that carry your identity. Your about page, your founder page, anything that states who you are and what you sell. AI systems read these to decide whether to name you.
• Pages holding something no model can generate. Your own test data, your own pricing, your own return numbers, your own photographs.

Now the harder call. Thin definitional content that ranked for years is the most exposed asset you own, and rewriting it rarely helps. If a 900 word post exists only to define a term, accept that a summary now does that job. Fold it into a deeper page carrying original information, or leave it in place and stop treating its sessions as a KPI.

What to change in your reporting

Reporting is where most teams get this wrong twice. They keep sessions as the headline number, then panic when it falls.

• Report clicks and impressions together, always. Impressions are now a visibility asset even when they do not become visits.
• Track branded search volume monthly as its own line. Unbranded informational traffic falling while branded search rises means your content is working through a channel your analytics cannot see.
• Segment organic reporting by query intent, not only by landing page.
• Set expectations with leadership in writing, before the quarter closes. Informational organic sessions falling 20 to 40 percent while revenue holds flat is not a failure. It is a channel changing shape.

The brands that handled this badly were the ones still optimising for a number that had quietly stopped meaning what it used to mean.