# AI Overviews traffic: search clicks are down 42%. Where does a small site get visitors now? > Search clicks are down about 42% and zero-click queries are near 69%. The channels that do not depend on a blue link, ranked by what they cost to try. Published 2026-09-05 · 5 min read · ai-overviews-traffic, seo, traffic Canonical: https://bidsurvivor.space/blog/where-small-sites-get-traffic-after-ai-overviews --- AI Overviews traffic loss is the defining distribution problem of 2026 for small sites, and the numbers are worse than the vibes. [Search Engine Land](https://searchengineland.com/google-ai-overviews-cut-search-clicks-report-471497) reports search clicks down roughly **42%** from baseline, a **61%** click-through drop on queries where an Overview appears, and zero-click behaviour at around **69%** of queries. If your plan was "rank, then wait", the plan now returns about two fifths of what it used to. This is not a post about whether that is fair. It is a list of where the remaining visitors actually come from, ordered by what each costs to try. ## First, work out how much of this is your problem Before rebuilding anything, separate your pages into the ones that face an Overview and the ones that do not. The 61% figure applies to the first group only. Informational queries — "what is X", "how does Y work" — attract Overviews. Comparison queries, queries with commercial intent, and long-tail specifics often still resolve to links. If most of your traffic came from the second group, you have a smaller problem than the headline suggests and should not panic-rebuild. If it came from the first, read on. ## The channels that do not need a blue link **Communities where your buyer already asks the question.** The highest-intent traffic available to a small site, and free apart from your time. The cost is that it cannot be automated and punishes anybody who arrives selling. Done properly it is [community listening](/blog/community-listening-find-the-thread) — reading for a week before posting once. **Launch platforms.** A concentrated burst rather than a channel, but the burst is real and the archive keeps sending a trickle. Worth doing once per meaningful release, not once a month. **Newsletters, other people's.** You are buying somebody else's relationship with their readers. Expensive per visitor compared with search, but the visitors arrive warm and no algorithm sits between you and them. **Being cited inside the answer.** If the Overview is eating the click, the remaining move is to be the source it quotes. That is a real, if unreliable, channel now — the mechanics are in [AI Overview citations](/blog/how-to-be-cited-by-an-ai-overview). **Bought placement on pages people chose to read.** Sponsorships, ad slots on sites with a real audience, and time-boxed placements. The advantage over search is that nothing between you and the reader can revise its click-through rate by 61% overnight; the disadvantage is that it stops when you stop paying. **Direct and returning visitors.** The only channel nobody can take away, and the slowest to build. Everything above is ultimately in service of this. ## What it costs to test each one The ranking that matters for a small site is not "which channel is biggest" but "which can I evaluate this week for near nothing". By that measure: - **Communities**: free, one week to a read. - **Being cited**: free, one afternoon of restructuring, unknowable payback. - **Launch platforms**: free, but you only get one shot per release. - **Ad slots**: cheap to a few hundred, depending on where. Same-day answer. - **Newsletters**: hundreds up, two to six weeks to a slot. Slow to evaluate. - **SEO**: the most expensive to test, because the test takes a quarter and the rules changed midway through your last one. That ordering is the actual argument of this post. When the reliable channel becomes unreliable, the correct response is not to pick a new reliable channel — there isn't one — it is to shorten your feedback loops so you find out faster which of these works for you specifically. Our own board is in the "ad slots" line and I will disclose the interest plainly: two twelve-hour slots a day, the first brand into an empty one pays nothing, every visitor starts with $100 of house credit, and the opens and click-throughs each slot produced are published on the front page and on each brand's page. It is a young board and honestly small; the reason to use it in this context is the same-day answer, not the volume. If you want the full priced comparison across channels, [the cheapest way to get website traffic](/blog/cheapest-way-to-get-website-traffic) sets them side by side per visitor. ## What not to do **Do not delete your content.** The pages still send the traffic they send. Deleting an underperforming asset because a headline scared you is how sites lose the 39% they still had. **Do not chase every new surface at once.** Discover, video, short-form and forums all get recommended in the same breath, and a one-person team running six channels badly does worse than one running two well. The [two channel rule](/blog/the-two-channel-rule) exists for exactly this moment. **Do not assume the 42% is permanent.** It is a measurement of a system that is being actively litigated and actively rewritten. Building an entire strategy on this quarter's coefficient is the same mistake as building one on the last quarter's. ## The uncomfortable summary Search was never a channel you owned; it was one you rented on unusually good terms, and the terms changed without notice. The lesson is not "SEO is dead" — that is wrong, and the sites saying it loudest are the ones with something else to sell you. The lesson is that a distribution plan with one input is fragile no matter how well that input has performed, and that the cheapest insurance is not a second channel but a **faster way of finding out** whether a channel works. Everything above is downstream of that.