“If AI gives away all the answers, who will ever click on our website?”
Since ChatGPT arrived and Google pushed AI Overviews to the top of the results page, that question has dominated marketing meetings. The assumption felt obvious: if the answer is already printed on the search page, nobody travels to the source. Google’s own data points the other way. In a post on the official Google blog, Liz Reid, VP of Google Search, reports that total organic search traffic to websites has remained stable since AI Overviews rolled out, while the character of each click has shifted toward what Google calls quality clicks.
That single sentence reframes the debate. The risk was never that clicks would vanish. The risk is that clicks would become cheap, arriving from users who wanted a headline and nothing more. Below are the three takeaways that matter for anyone running a website, each paired with the action to take this quarter.
💡 Point 1: AI Doesn’t Steal Traffic; It Generates ‘Quality Clicks’
The most reassuring part of the announcement concerns volume. VP Liz Reid states plainly that “the total volume of organic search traffic to websites has remained stable even after the introduction of AI Overviews.” Aggregate traffic held. What changed is the intent behind each visit.
In the classic search model, a user scanned ten blue links, picked the most clickable title, landed on the page, and often bounced within seconds because the snippet had already answered the question. AI Overviews invert that sequence. The summary resolves the surface-level question on the results page, so the people who still click are the ones who want depth: pricing tables, real photographs, process details, a named expert’s opinion. Google calls these quality clicks, and for site owners they translate into lower bounce rates, longer dwell time, and a higher share of visitors who are genuinely evaluating a purchase.
| Dimension | Pre-AI Overviews clicks | Post-AI Overviews clicks |
|---|---|---|
| Trigger for the click | Curiosity, catchy title | Unresolved need, need for proof |
| Typical session length | Short, single-page | Longer, multi-page exploration |
| Bounce rate pattern | High | Lower |
| Conversion proximity | Mostly top of funnel | Mixed, with more mid and bottom funnel |
| What wins the click | Headline craft | Evidence, specificity, first-hand detail |
One caveat worth stating honestly: click-through rate per impression has fallen. Pew Research Center found in a July 2025 analysis that users who encounter an AI summary are markedly less likely to click a traditional link than users who do not. Both findings can be true at once, because query volume grew at the same time. Fewer clicks per search, more searches overall, and better clicks per visit.
Practical move: stop reporting raw sessions as your headline metric. In GA4, build an exploration that compares engaged sessions per click and average engagement time for landing pages that appear as AI Overview sources against those that do not. If the cited pages show longer engagement but fewer sessions, you are seeing the shift in action, and the correct response is to deepen those pages rather than chase volume.
💡 Point 2: People’s ‘Queries’ are Evolving
The second change is behavioral. Once users learned the model can hold context, they stopped compressing their needs into two-word keyword strings. The old query was “Hongdae restaurants.” The new query is “recommend a vegan restaurant in Hongdae with parking where ten people can have a quiet conversation.”
This is not a cosmetic difference. A conversational query carries constraints, and constraints are where commercial intent lives. It also expands the total search pie, because people now ask things they previously considered too awkward or too long to type. Google’s post frames this as more queries overall, and that framing matters for planning: the opportunity is not to defend a shrinking set of head terms, but to occupy a growing set of specific questions.
| Comparison | Short-tail keyword | Conversational long-tail query |
|---|---|---|
| Example | Hongdae restaurants | Vegan restaurant in Hongdae with parking for 10 people |
| Competition | Extremely high, dominated by aggregators | Lower, often decided by one well-matched page |
| Intent clarity | Ambiguous | Explicit constraints and context |
| Content that wins | Broad category pages | Specific answers with conditions addressed |
| AI Overview citation odds | Low, generic sources preferred | Higher, specific sources are needed |
Practical move: mine the question layer instead of the keyword layer. Pull the “People also ask” expansions for your core topics, export your site search logs, and read the last twenty inbound emails or chat messages. Each real question becomes an H2 written in the user’s own phrasing, answered in the first two sentences, then supported with detail. Pages structured as a chain of answered questions are far easier for an AI system to quote than a single 2,000-word essay with one vague heading. If you want the markup-level mechanics behind that structure, our guide to content SEO with heading tags and internal links covers how H2 and H3 layers should be arranged and how each section should connect to the next.
💡 Point 3: In the Age of AI, ‘Human Experience’ is More Valuable Than Ever
So which pages does Google choose to cite inside an AI Overview? The post is explicit that “users still want the unique perspectives and authentic experiences of others.” When a machine can compress universally known facts into three lines, a page that only restates those facts has nothing left to offer. The destinations that keep earning traffic are the ones carrying something the model cannot generate: a measurement someone actually took, a failure someone actually survived, a photograph taken on site last Tuesday.
Field note from a real operator. A small clinic owner we spoke with rebuilt a single treatment page after watching AI Overviews absorb her generic intro copy. She added the actual equipment model and purchase year, the number of procedures performed in the previous twelve months, a photo of the recovery room, and a short paragraph on the two cases she now declines to treat. Sessions to that page dropped by roughly a fifth over the following quarter, while booking inquiries from the same page rose. Fewer visitors, better visitors, and a page that finally reads like it was written by a person who was there.
Practical move: audit your five highest-value pages and ask a blunt question about each one. Could a competent writer with no access to your business have produced this text from public sources alone? If yes, the page is replaceable. Then add at least three elements that require insider access: a dated first-hand observation, a specific number from your own records, and a limitation or trade-off you are willing to admit publicly. Admitting a limitation is the fastest credibility signal available, and it is precisely the kind of nuance a summary cannot invent.
📝 Editor’s View: Not the End of SEO, But a ‘New Evolution’
Read the whole post and one conclusion holds: generative search is not the enemy of websites. It behaves like a demanding curator that routes people toward material worth their time. The uncomfortable part is that the curator has no patience for filler.
Outdated tactics such as keyword padding, thin aggregation, and pages assembled from other pages are losing value quickly. What gains value is the depth and originality with which a site answers its customers’ real, complicated questions. Sharpening the authentic experience and demonstrable expertise that AI cannot imitate is the most durable marketing position available in this cycle, and it is the direction Google itself points to. For a wider playbook that pairs conventional search optimization with answer-engine visibility, see our breakdown of the AI search era SEO and GEO integration strategy.
❓ Quick Q&A on AI Overviews and Traffic
Q. Did AI Overviews actually reduce website traffic?
A. Not in aggregate. Google reports total organic search traffic to sites has stayed stable, though click-through rate per impression has declined as query volume grew.
Q. What exactly is a quality click?
A. A visit from someone whose basic question was already answered on the results page, so they arrive wanting depth, proof, or a decision rather than a definition.
Q. Which pages are most likely to be cited as sources?
A. Pages with first-hand detail, original data, named authorship, and clear question-based headings. Generic summaries of public facts are the least likely candidates.
Q. How should keyword planning change?
A. Shift from head terms to question clusters. Build one page per real question, answer it in the opening two sentences, then support it with specifics that only your business can supply.
📌 Read the original article here: Google Official Blog – AI Search driving more queries, higher quality clicks
📊 Referenced data: Pew Research Center (July 2025) – Google users are less likely to click on links when an AI summary appears in the results