By Anthony Neal Macri
Organic search traffic is sliding almost everywhere, and most of the worry about it is aimed at the wrong target. The thing that should keep you up isn’t how much traffic is left. It’s whether an AI names you when it answers the question.
Every few weeks now I get a version of the same email. Traffic’s down, the client says. Sometimes there’s a screenshot attached. Organic sessions are off twenty percent against last year, sometimes thirty, once close to forty, and underneath it the same question every time: what did we break?
Almost always, the honest answer is nothing. I go and look anyway, because you have to, and the site is fine. Rankings holding. Content solid. No penalty, no technical mess buried in the code. That’s the part people struggle with, because a clean audit and a falling graph aren’t supposed to sit next to each other. For twenty years they didn’t.
The arrangement between a company and a search engine used to be almost embarrassingly simple. You published something worth reading, Google filed it away, and when somebody searched, they landed on your page. Traffic was the whole point, and a profession grew up around earning more of it. What’s happening now isn’t a tweak to that deal. It’s the deal ending, replaced by a results page that answers the question itself and lets the user get on with their day.
The number that hides the bodies
The top-line figure doesn’t look alarming, which is half the problem. U.S. organic search traffic was down roughly 2.5 percent year over year heading into 2026, going by figures reported in Search Engine Land. You could live with 2.5 percent. But an average that mild is hiding a lot underneath it.
Watch what happens when Google runs one of its AI Overviews, the synthesized answer that now sits on top of the old blue links. Seer Interactive looked at millions of queries and found organic click-through rate falling from about 1.76 percent to 0.61 percent when an Overview was present. Call it a sixty percent drop. Other studies come at it from different directions and end up in the same neighborhood: most searches now finish without a single click, and on the queries where an AI answer appears, the no-click rate runs north of eighty percent.
It doesn’t land evenly, either. The queries that got hollowed out are the ones an AI can dispatch in a sentence. Definitions. How-tos. “What’s the best way to.” Break the damage out by sector, and it gets bleak fast wherever those Overviews show up most: health sites down about a third year over year, personal finance not far behind, travel close after that. The pages that held their ground, or grew, tend to be the ones the AI actually pulls from.
Then there’s the public wreckage. HubSpot was reported to have shed something like seventy to eighty percent of its organic traffic, with its own chief executive more or less telling investors that AI answers were satisfying people upstream of the click. Business Insider disclosed a decline of around fifty-five percent over three years and cut staff afterward. These weren’t amateurs running thin content. They were serious, well-resourced content operations, and it happened to them regardless.
If total sessions are your only scoreboard, this is a catastrophe. But no board ever actually asked for pageviews. They asked for pipeline. And that is exactly where the story stops being about decline.
Fewer visitors, and better ones
The visitors who do still arrive from AI search convert like nothing organic ever did.
The studies stack up through 2025 and 2026 and keep landing in the same range: AI-referred visitors converting somewhere around four to five times the rate of ordinary organic, with the gap widening in B2B and professional services. Seer clocked ChatGPT referrals converting near sixteen percent, against under two for Google organic. Ahrefs found a sliver of AI traffic responsible for a wildly outsized share of its signups. Adobe watched AI-referred shoppers outconvert every other channel in retail, having seen that same channel lag a year earlier.
Once you see why, it stops being surprising. When somebody asks an AI which vendor to consider, the model has already done the legwork. It read across the field, weighed the options, and put your name forward. The person who clicks that citation isn’t starting their research. They’re finishing it. They turn up warm, late in the process, already half-sold by a source they trusted enough to ask in the first place. No keyword ever delivered you a lead in that condition.
Which is why the question worth asking has quietly changed.
You’re not tuning a website. You’re building a consensus.
For twenty years the prize was a position on a page. Now it’s a place in the answer. A ranking got you into a list; a citation gets you named in the sentence a person reads. And the list, of course, is the very thing people have stopped scrolling.
So why does an engine cite one company and skip past the next? Not for the reasons the old playbook drummed into everyone. Some of those reflexes actively hurt you now. Stuffing keywords, for one, has been shown to push generative visibility down rather than up. What gets you cited looks a great deal more like an actual reputation.
It starts with authority, and by that I mean what the rest of the web says about you, not what you say about yourself. These systems lean hard on outside corroboration. A large share of the brand mentions that surface in AI answers trace back to earned media and independent sources rather than a company’s own domain. Concrete numbers help, too; optimization studies found that pages carrying real data and citing credible sources were roughly forty percent likelier to get quoted. Structure counts, because the model has to lift a clean answer out of your page without a struggle. Freshness counts, because stale pages quietly fall out of rotation. And here is the one I’d underline for any executive in the room: when several independent sources describe a company the same way, the models begin treating that description as fact. The web’s rough consensus about you becomes the machine’s verdict on you.
That’s the whole game. You’re not really tuning a website anymore. You’re shaping a consensus and hoping it’s the one you’d have chosen.
I’ve watched this run both directions. One client of mine, a medical-device distributor in a heavily regulated corner of the market, kept its organic footprint more or less flat through 2026 while comparable names drifted down with the broader slide. There was no clever trick behind it. We’d spent months laying down precisely what an engine needs to cite you without hesitating: a proper clinical resource library, content shaped so an answer can be pulled straight out of it, entity details that stay consistent everywhere they appear, and a steady flow of outside coverage, awards, trade press, and the occasional independent write-up. When the engines went looking for a credible source in that category, one was already sitting there.
A software company I work with in Europe took it further still. We more or less stopped fussing over rankings and started building a citation base instead. Founder thought leadership, speaking slots, bylines in the trade press. Almost none of it lives on the company’s own website, which is the entire point. The asset that compounds isn’t the site. It’s the accumulated agreement out there about who these people are and what they’re good at. That’s what the AI reads when it decides whether to say their name.
That reframe is what I build engagements around now. I’ve taken to calling it a Growth Visibility Framework, though the label matters far less than the shift underneath it: stop working to be found, and start working to be quoted.
Why I’d move on this first
Here’s the part I’d act on before anything else if I were running a marketing team this year. Authority of this kind is slow. You can’t buy it in a quarter, and you can’t fake it convincingly for long. But once it’s real, it compounds, and it’s stubborn to dislodge. That combination is what makes right now a window, and windows close.
The citation game is already crowded and already twitchy. Cited sources reshuffle month to month; lose your spot, and it can take weeks to claw back, usually because a competitor slipped into it while you weren’t watching. The companies putting genuine effort into authority now aren’t only defending a channel. They’re claiming the default answer for their whole category, the thing an AI says about their industry before a future customer has heard of any of them. Own that answer, and you own the introduction.
So I wouldn’t spend this year’s budget chasing back clicks that aren’t coming back. I’d spend it making sure that when an AI speaks for my industry, it knows my name and offers it up with confidence. That’s the introduction that matters now, and most of your competitors haven’t clocked it yet. For a little while longer, that gap is the whole opportunity.
Anthony Neal Macri
