Google's AI Overviews Are Cutting Your Clicks - What the Data Says and What Actually Helps

Type a question into Google today, and there's a good chance you'll never click a single link. An AI-generated summary  - Google's AI Overview  -  sits above the results, hands the searcher an answer, and often ends the search before it starts.

For years, "rank #1" was the entire SEO strategy. That advice is now incomplete. Ranking well still matters  -  it's still the raw material AI Overviews draw from  -  but it no longer guarantees a click. Multiple independent studies published between mid-2025 and mid-2026 now put numbers on exactly how much that's changed, and the picture is consistent even where the exact figures differ.

What Are Google AI Overviews, Exactly?

AI Overviews are AI-generated summaries that appear above traditional search results. They pull information from multiple websites, synthesize it into a single answer, and cite a handful of sources  -  usually three or more  -  beneath it. The searcher often gets what they need without clicking through anywhere.

This is a structurally different mechanism than the old featured snippet, which lifted one passage from a single page. An AI Overview blends several sources into one narrative, which means your content can shape the answer without your brand getting the click, or sometimes even the visible credit. That gap  -  your expertise powering an answer that someone else's link sits under, or that no one clicks at all  -  is the core tension behind everything else in this piece.

How AI Overviews Actually Decide What to Cite

Google's system doesn't rank pages the way traditional search does. In broad strokes, it:

1. Identifies the underlying question, even when the query is phrased conversationally.
2. Pulls from a pool of pages that already rank reasonably well in traditional search.
3. Extracts short, self-contained answer segments that directly resolve the question.
4. Synthesizes multiple segments into one coherent summary.
5. Attaches citations to a handful of the sources it drew from.

Pages that get pulled into this pool tend to share traits: a clean structural hierarchy, an answer that appears early without requiring inference, and clearly stated supporting facts. Pages that bury the answer under long preambles are less likely to be extracted cleanly, even if they rank well.

The Data: What's Actually Happening to Clicks

Numbers on this vary by methodology, panel, and time period  -  that variation is itself worth noting rather than smoothing over. Here's what the more rigorous, named studies report:

Two things are worth taking from this rather than any single headline percentage: the direction is unambiguous across every methodology, and the size of the drop is genuinely a range (roughly 35%–65% depending on study and time window), not a fixed number. Anyone citing one figure as the number is oversimplifying. It's also worth noting Seer's data shows a partial rebound in early 2026 as Google reworked how it displays links inside summaries  -  this is a moving target, not a settled state.

Who Gets Hit Hardest

• Small businesses and niche publishers feel it most, since their traffic leans heavily on unbranded, informational queries  -  exactly what AI Overviews are built to answer directly.
• Large, recognizable brands are more insulated, because a meaningful share of their traffic is branded search (people searching the company name), which AI Overviews rarely displace.
• Transactional and commercial-intent searches ("best," "buy," "pricing," comparisons) hold up better than purely informational ones, since an AI summary can't complete a purchase for the user.

What's Actually Working - and What Isn't

This is the part where a lot of advice gets ahead of the evidence. Some tactics have real support; others are weaker than commonly claimed.

Reasonably well-supported:

• Answer-first structure. Pages with a direct, self-contained answer near the top of a section are mechanically easier for the extraction process described above to lift cleanly.
• Original data and first-party research. Content that's the origin of a fact  -  an internal survey, a proprietary benchmark  -  has a structural advantage over content that just repeats what's published elsewhere, since AI systems are drawing from a pool of similar pages and something has to differentiate the source.
• Being cited pays off, modestly. Seer Interactive's data found brands cited inside an AI Overview see meaningfully higher organic CTR than brands whose content feeds the summary without a visible citation  -  being named as the source, even with lower absolute click volume, functions more like brand placement than like a traditional referral.

Weaker than the conventional wisdom suggests:

• Schema markup, on its own, is not a reliable lever. An Ahrefs study covering 1,885 pages that added JSON-LD schema between August 2025 and March 2026  -  measured against 4,000 matched control pages  -  found no statistically significant citation increase on Google AI Overviews, AI Mode, or ChatGPT. A separate industry test at a March 2026 conference reported large citation gains from a sitewide schema rollout, so the evidence is genuinely mixed. The more defensible read: schema helps AI systems correctly parse and disambiguate content that's already strong, and may help a brand-new page enter the citation pool faster, but it doesn't rescue weak content and shouldn't be treated as a guaranteed lever on its own.
• FAQ blocks added purely for schema's sake. Multiple sources note that formulaic Q&A sections stapled onto pages without genuine relevance underperform Q&A content that reflects real user questions.

The 2026 Playbook

1. Write answer-first, then expand. Lead each section with a direct, extractable answer in one or two sentences, then add supporting detail and nuance.
2. Invest in genuine E-E-A-T signals. Named authors with real credentials, first-hand case studies, and transparent sourcing matter more, not less, when an algorithm is choosing what to trust.
3. Use structured data as a supporting signal, not a silver bullet. Add FAQ, HowTo, and Article schema where it reflects real content  -  but don't expect it alone to move citation rates, per the mixed evidence above.
4. Structure content around real user questions. Genuine Q&A formatting is one of the higher-leverage, better-supported tactics here.
5. Double down on branded search. Since branded queries are largely AI-proof, PR, podcast appearances, and community presence that get people searching your name directly is a durable hedge.
6. Rebalance investment toward transactional content. Purely informational content is the most exposed category; comparison and product pages are less replaceable by an AI summary.
7. Diversify traffic channels. Email, communities, social, and direct traffic all buffer against continued volatility in how search results get consumed.
8. Publish original data. Internal surveys, proprietary benchmarks, and first-party case studies give AI systems a reason to treat you as the origin rather than one of several interchangeable sources.
9. Track citations, not just rankings. A page can hold position perfectly and still lose most of its clicks. Google's Generative AI performance report (added to Search Console in mid-2026) is a starting point, though it has real limitations worth checking before drawing conclusions from it.
10. Keep the technical fundamentals clean. Slow pages, intrusive pop-ups, and poor mobile rendering hurt both traditional crawlability and AI extraction alike.

A Step-by-Step Implementation Checklist

1. Audit your top 20–50 informational pages for current AI Overview appearance, manually or with a tracking tool.
2. Identify the core question each page answers, and check whether that answer appears in the first 100 words.
3. Rewrite section openings to lead with a direct answer before expanding into detail.
4. Add schema markup where it reflects genuine content structure  -  treat it as hygiene, not a growth lever on its own.
5. Insert at least one original data point or first-hand insight per page, even something as simple as a small internal survey.
6. Break long paragraphs into scannable chunks with subheadings that mirror how people actually phrase questions.
7. Re-test AI Overview appearance for the same queries four to eight weeks after changes go live  -  most studies report that as the window where changes become measurable  -  and iterate.
8. Expand investment in branded and transactional content to reduce dependence on informational-query traffic specifically.

    Traditional SEO vs. AEO/GEO-Optimized Strategy

Common Mistakes

• Publishing more content instead of restructuring existing content. Volume rarely fixes an extraction problem  -  clarity does.
• Treating schema as a guaranteed fix. The evidence is genuinely mixed; don't oversell it internally or to clients.
• Stuffing FAQ blocks onto every page regardless of relevance. Genuinely useful Q&A outperforms formulaic filler.
• Ignoring branded search. Brand-building is a traffic-diversification strategy now, not just a marketing nicety.
• Measuring success by rank alone. A page can hold position and still lose most of its clicks  -  track citations too.

The Bottom Line

The data is now specific enough to say this plainly: AI Overviews have measurably cut organic clicks, by roughly a third to two-thirds depending on the study, and that's not a temporary blip  -  Seer's own longitudinal data shows a partial rebound, not a return to pre-AI-Overview levels. The tactics that hold up under scrutiny are the boring, defensible ones: answer-first writing, real original data, genuine expertise, and less dependence on any single channel. The tactics that don't hold up as well  -  schema as a cure-all being the clearest example  -  are worth dropping from the pitch deck even if they stay in the technical checklist.

Frequently Asked Questions

1. What are Google AI Overviews?

AI-generated summaries that appear at the top of search results, synthesizing information from multiple websites into a single answer, with source links shown below.

2. Why are AI Overviews reducing website traffic?

Because they often answer the question directly on the results page. Pew Research's tracking of real user behavior found organic click rates roughly halve  -  from 15% to 8%  -  when an AI summary appears, and only 1% of visits click a link inside the summary itself.

3. Do AI Overviews affect all searches equally?

No. Informational queries (how-to, what-is, comparisons) are hit hardest. Transactional searches (buying, pricing, booking) hold up better, since an AI summary can't complete a purchase.

4. Does schema markup guarantee inclusion in an AI Overview?

No  -  and the evidence for it as a strong lever at all is more mixed than commonly claimed. An Ahrefs study of 1,885 pages found no significant citation lift from adding schema alone. It likely still helps disambiguate content and may help new pages enter the citation pool, but it isn't the guaranteed win it's often sold as.

5. Is traditional SEO still relevant?

Yes, but it's necessary rather than sufficient. Ranking well is still required to be eligible for citation  -  AI Overviews pull from pages that already perform reasonably well in traditional search  -  but ranking alone no longer guarantees traffic.

6. What's the difference between SEO, AEO, and GEO?

SEO focuses on ranking in traditional search results. AEO (Answer Engine Optimization) focuses on getting content selected as a direct answer. GEO (Generative Engine Optimization) focuses on getting cited within AI-generated responses across Google, ChatGPT, and Perplexity. Most brands now need to address all three.

7. Should small businesses stop investing in informational content?

No. It still builds topical authority and remains the source material AI Overviews draw from. The better move is restructuring it for extractability while also building out branded and transactional content, rather than abandoning it.