Ask most publishers what's killing their search traffic and you'll get the same answer: AI Overviews, AI Mode, chatbots eating the clicks that used to land on their site. Chartbeat's own data says something more specific, and more uncomfortable for smaller operations: the traffic loss isn't evenly distributed, and the thing it correlates with most cleanly isn't AI usage. It's size.
The size gap, in numbers
Chartbeat measured search referral traffic across its publisher network from December 2024 to December 2025, segmented by daily pageview volume. The pattern holds cleanly across three tiers.
| PUBLISHER SIZE (DAILY PAGEVIEWS) | SEARCH REFERRAL DECLINE | RELATIVE TO LARGE SITES |
|---|---|---|
| Small (1,000-10,000) | -60% | 2.7x worse |
| Medium (10,001-100,000) | -47% | 2.1x worse |
| Large (100,000+) | -22% | baseline |
This tracks closely with independent reporting on the same underlying data: Axios ran it as an exclusive, and Search Engine Journal covered the same figures shortly after. Google Search pageviews across Chartbeat's network fell 34% between December 2024 and December 2025 overall, and Google Discover fell 16% over the same window, before you even split by publisher size. The size breakdown is what turns a bad-year-for-everyone story into something with an actual mechanism worth understanding.
It's worth sitting with how wide that gap actually is before moving on. A 2.7x difference in decline rate between the smallest and largest tier isn't a rounding error in how Chartbeat bucketed its data. It's the difference between a publisher that's bruised and a publisher that's bleeding out. A large site losing 22% of search referral traffic is a bad year that a diversified revenue base can absorb. A small site losing 60% of the same channel, when search referral was likely the majority of its total traffic to begin with, is closer to an existential threat than a rough patch.
The three-tier structure Chartbeat used, 1,000-10,000, 10,001-100,000, and 100,000-plus daily pageviews, roughly maps onto the shape of the publisher world itself: small tends to mean a niche blog, newsletter-adjacent site, or single-topic trade publication; medium tends to mean a regional outlet or a mid-size vertical media brand; large tends to mean a national or global publisher with a diversified editorial staff and, usually, a diversified traffic base already. The decline curve tracking that size ladder so cleanly is itself evidence that whatever mechanism is driving this, it correlates with structural publisher characteristics, not with how good any individual publisher's content actually is.
Why AI chatbots aren't the culprit
Here's the finding that should reset a lot of publisher anxiety, even if it's not a comfortable one: AI chatbots, ChatGPT, Perplexity, Claude, and the rest, account for less than 1% of total pageviews across Chartbeat's entire network. That's total pageviews, not just search-referred ones. Whatever is driving a 60% search referral collapse for small publishers, it is not primarily traffic getting siphoned off to AI chat interfaces instead.
That doesn't mean AI is irrelevant to the story. ChatGPT referrals specifically grew more than 200% year over year in the same dataset, a real and fast-growing number. It's just growing from a genuinely tiny base, and even after doubling several times over, it remains a rounding error next to the scale of what's been lost from classic search referral. A publisher chasing the AI-traffic narrative as the explanation for a 60% search decline is chasing a signal that, per this data, explains almost none of the actual loss.
This is worth stating plainly because the incentives point the other way. "AI is stealing our traffic" is a more satisfying explanation than "Google's algorithm has started favoring the sites that already had the most authority," for reasons that have nothing to do with which explanation the data actually supports. The first one names an external villain and implies a specific, tractable fix: get cited by the AI tools instead of losing to them. The second one is a structural, compounding disadvantage that doesn't have a clean single fix, and it implicates the same search relationship publishers have depended on for two decades, which is a much harder story to tell a newsroom or a board.
What's actually driving the split
If it isn't AI chatbots pulling traffic away, the size-correlated pattern points somewhere else: how Google's evolving results, AI Overviews included, allocate the shrinking pool of clicks that search still sends out. When an AI Overview appears, it compresses the space available below it, and whatever the algorithm decides deserves the remaining prominent placement tends to concentrate around sites Google already treats as high-authority, high-trust, high-engagement, exactly the profile large publishers have by default and small ones have to fight for.
A large site with deep archives, high direct and returning-visitor traffic, and established topical authority has more signal for Google to lean on when deciding who still deserves visibility as the visible SERP real estate shrinks. A small site, even one publishing genuinely good, accurate, well-researched content, has less of that accumulated signal to fall back on, and it shows up disproportionately in exactly the tier that lost the most.
Think about what Google's ranking and AI-answer systems actually have to work with when they're deciding whom to favor as available space contracts. Engagement history, return-visitor rate, dwell time, brand-search volume, backlink profile depth, all of these accumulate over years and scale with audience size in a way that's structurally hard for a smaller, newer, or more focused publication to match regardless of how good any individual piece of its content is. A shrinking pool of visible placements gets allocated according to those accumulated signals first, and content quality on any single article is, at best, a tiebreaker within a tier the site already qualifies for.
This is consistent with a pattern we've tracked elsewhere: freshness and authority signals increasingly favor sites that already have a track record, not just the site with the technically best answer to a given query today. Topical authority compounds for sites that already have a lot of it, and the compounding effect is exactly what would produce a decline curve that gets steeper the smaller the publisher.
There's a useful analogy in how index funds behave during a market selloff. Large, liquid, well-covered names get bid back up faster because there's more institutional confidence and more existing capital already positioned there. Smaller, thinner names take longer to recover and sometimes don't. Google's algorithmic allocation of a shrinking pool of search placements is behaving similarly: the sites with the deepest existing reserve of trust signal recover their share of the shrinking pie faster, or lose less of it in the first place, than sites with a thinner reserve to draw on. It's not a conspiracy against small publishers specifically. It's what naturally happens when a system built on accumulated signal gets forced to allocate scarcity.
What smaller sites can do about it
The instinct to fight this by chasing AI Overview citations specifically is, per this data, solving for the wrong 1%. The bigger lever for a small or mid-size publisher is the same one that's always mattered, just with higher stakes now that the margin for being merely adequate has shrunk: build genuine topical depth in a narrow enough lane that authority actually compounds, rather than spreading coverage thin across a wide beat competing directly with sites ten times the size.
Narrowing the beat is the least intuitive part of that advice for a small newsroom used to thinking growth means covering more, not less. But the mechanism this data implies rewards depth within a lane over breadth across one, because depth is exactly what accumulates into the kind of authority signal a large publisher already has by default. A small site that publishes broadly across ten loosely related topics never builds enough signal in any one of them to compete for the shrinking pool of visibility in that specific category. The same publishing effort, concentrated into two or three topics the site can genuinely own, has a real shot at building the kind of compounding authority that shows up favorably in exactly the systems currently squeezing everyone else out.
Diversify the traffic mix deliberately rather than treating search as the default channel and everything else as a bonus. Chartbeat's broader point, echoed across the reporting on this data, is that the publishers holding up best are finding real traffic through direct visits, email, apps, and returning-visitor loyalty, channels a search algorithm doesn't gate. That's not a new idea, but this data makes the case for prioritizing it now rather than later with actual numbers instead of just industry anxiety.
Email specifically deserves more weight in that mix than most small publishers currently give it. A newsletter list is one of the few audience assets a search algorithm can't touch, can't reweight, and can't deprioritize based on a site's overall authority score. It's slower to build than search traffic ever was, and it doesn't scale the way a viral search hit can, but it's durable in exactly the way search referral just proved itself not to be for smaller publishers, and durability is the thing this data says is actually scarce right now.
It's also worth being honest that not every small publisher can out-execute this gap through content strategy alone. Some of what's happening here is a structural feature of how algorithmic ranking systems behave under scarcity, not a solvable editorial problem, and a realistic strategy has to include revenue diversification, not just traffic diversification: memberships, sponsorships, and direct reader relationships that don't route through a search algorithm's evolving judgment about who deserves the remaining clicks at all.
And resist the urge to blame every traffic dip on AI specifically, because it distorts what actually needs fixing. A content strategy response built around "get cited by ChatGPT more" doesn't touch the 99% of the problem this data says is coming from somewhere else entirely: how Google is currently allocating a shrinking pool of clicks across sites of different sizes and authority levels, a technical SEO and authority-building problem that predates AI Overviews and will outlast the current AI-anxiety news cycle. Chartbeat's own writeup on the full dataset is worth reading directly for the size breakdown by month rather than just the year-over-year headline: Pageviews are down, but AI's impact is complicated.
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Josh leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.