The mistake most teams make about AI Overviews is treating them as one uniform hit. They are not. Independent research — Seer Interactive and Ahrefs studies, plus a randomized field experiment from Carnegie Mellon and the Indian School of Business — consistently finds that the damage depends almost entirely on the intent behind the query. Split your keyword set by intent and the confusing aggregate number resolves into four very different situations, each with its own correct move. This is the matrix, built from the segment data and cross-checked across the January-to-September 2026 evidence.
The four intents, and what each one costs:
- Informational — the critical risk. Queries that trigger an AI Overview lose roughly 70 to 80% of their clicks. This is the segment the answer surface eats.
- Commercial — moderate. These lose roughly 35 to 45% of clicks.
- Transactional — the holdout. These lose the least, around 20 to 30%.
- Branded — the surprise winner. Branded queries actually gain about 18.7% click-through when an overview appears.
Scattered across those four is the single most useful fact in the whole dataset: brands that are cited inside the overview get 35% more organic clicks and 91% more paid clicks than brands left out. Each segment has a move that leans on that fact.
Informational: defend by earning the citation
The 70 to 80% loss is where the overview is doing the reading for the user, so ranking alone will not save you — the answer is a summary, and the user reads it and leaves. The only thing that recovers value here is being the source the summary cites. Build the answerable page: a clear, verifiable claim in the first paragraph, real data and named definitions, and enough structure for an engine to quote. This is also where the cited-brand advantage is at its most valuable — a cited brand in informational can still capture the 35% organic / 91% paid uplift while an uncited competitor captures nothing. If a query in this segment is purely informational and you can realistically win the citation, defend it. If you cannot, this is the segment to stop spending against and let go, because a blue-link number in this segment is worth very little.
Commercial: defend the page, and consider paid
At 35 to 45% loss, commercial is still winnable, but the balance shifts. This is the segment where users are comparing — business software, a service, a product decision — and where the engine is weighing sources rather than simply summarizing. Comparison pages and clearly scoped service pages are exactly the assets that get cited, so this is the best place to spend the "build the citable page" budget. And because the user is closer to a decision, this is also the segment where a paid slot on your organic gap matters: if the overview favors a competitor, a paid placement can still put you in front of the user. For commercial queries, the defensible plan is two-sided — strengthen the citable comparison page on organic, and reserve paid on the specifically contested queries.
Transactional: hold the line, keep it simple
At 20 to 30%, transactional queries hold up best, and the winning move is the least dramatic. Do not over-optimize this segment for AI citation — the user wants to act, and a clean product or pricing page that loads fast and makes the action obvious is the asset that converts here. Keep your transactional pages lean and fast, keep structured data clean so the click is natural to take, and prioritize this segment for CRO rather than for GEO. The 20 to 30% remaining gap is mostly a matter of capturing the users who still click, and nothing wins those back like a page that resolves the intent in one move.
Branded: protect and amplify
Branded is the one segment that gains, and it is easy to squander. The 18.7% click-through lift on branded queries is a signal that the overview is reinforcing brand recognition — protect it and lean into it. Keep branded query health clean (consistent entity data, a strong brand page, no confusing duplicates), and make sure your branded snippets read well. This is also the cheapest place to build the buffer that absorbs losses elsewhere: the Daily Mail pattern shows that strong branded and direct traffic is what lets a publisher take an 80 to 90% click-through crash on a keyword with almost no total-traffic impact. Brand the product, grow direct habits, and you are buying resilience against the segments you cannot fully defend.
The matrix in practice
The usable output is not a report, it is a decision. Pull your keyword set, tag each query with one of the four intents, and apply the matrix:
- Informational — win-by-citation or let go; do not spend on a blue-link number.
- Commercial — build the citable comparison page and reserve paid on the specific gaps.
- Transactional — keep it lean and convert; deprioritize GEO on this segment.
- Branded — protect and amplify; it is your cheapest resilience.
Then re-base how you report it. A leader who sees "clicks down 50%" is guessing, but the same leader who sees "we lost 70% in informational, gained in branded, and hold on transactional" knows exactly where to act — and where to stop acting. That granularity, not the aggregate, is the actual source of a defensible AI Overview strategy.
The honest limits
The segment deltas come from vendor and working-paper research, and the exact percentages for your site will differ. They are reliable as a direction and a prioritization, not as a guarantee — use them to decide where to spend, then verify against your own GA4 and Search Console numbers, split by intent, over the next two to three months. One caveat: the branded-uplift figure is the least robust of the four, so treat it as directional rather than a budget line. What is solid is the shape: informational is the hit, branded is the win, and your strategy should be built around taking the delta.
FAQ
Why are informational queries hit hardest?
Because the overview answers an informational question entirely — the user reads the summary and leaves, so there is nothing for a blue link to win. Commercial and transactional queries still require a choice or an action, which is why they retain more of their clicks.
Should I abandon my informational content?
Not necessarily — but change what you expect from it. If you can win the citation, keep optimizing the page for that and the branded uplifts can carry it. If you cannot realistically be cited, treat that content as brand and authority support and stop measuring it in clicks, because a blue-link number in informational is mostly an echo of a surface that no longer pays.
How do I tag my keywords by intent?
Use the main search intent from any keyword research too, or classify by hand from the query language — "best", "vs", "cost", "compare" lean commercial; "how to", "what is", "why" lean informational; "buy", "pricing", "discount", "near me" lean transactional; and queries containing your brand name are branded. Then apply the segment deltas to your own volume.
Related reading
- How to Stay Visible in Google AI Overviews — the seven-step tactical playbook
- AI Overviews Are Cutting Clicks by Up to 61% — how to defend your SEO strategy and budget
- AI Search Citation Sources by Industry — the citation base AI answers lean on
Author: Elena Shaw, AI Search Visibility Analyst at Auspia. Elena tracks how answer engines change the way content gets discovered and cited.
