Ryan Law at Ahrefs published their 2026 AI marketing trends this morning. Eight trends, backed by Keywords Explorer and an analytics cohort of roughly 75,000 sites. It is the best-sourced read on this I have seen this year and you should go read it.
I want to take two of his numbers and put a client’s Search Console underneath them, because I have been living inside that account for a month and it says the same thing from the other side.
Here is the first. Search demand for AI visibility tooling, year over year:
Generative engine optimization: 7,200 a month, up 84%
AI visibility: 4,400 a month, up 145%
AI visibility tracker: 2,400 a month, up 156%
AI visibility tools: 2,100 a month, up 266%
LLM visibility: 1,200 a month, up 126%
And here is the second, from the same publication on the same day. ChatGPT’s referral share across those 75,000 sites: 0.32% of all traffic.
An entire category of tooling, pricing, and agency service lines is being built on a channel delivering one third of one percent.
I sell AI visibility work. Weigh that accordingly, and keep weighing it, because what follows argues against the version of my own service that most people are currently buying.
The category is smaller than it feels
Law is right that this is a real emerging channel and the growth is genuine. Here is what I would add from sitting on the sell side of it.
Add up that whole keyword cluster and you get roughly 17,000 monthly searches. All of it. Every term in the category. “AI image generator” is 361,000 a month by itself.
The growth rates are real. The absolute numbers are small, and percentage growth off a base of near zero always looks like a revolution. “AI visibility tools” going up 266% means it went from very small to slightly less small.
The part I would push on is who those searches belong to. Ask yourself who types “AI visibility tracker” into Google. It is rarely a business owner wondering whether customers can find them. It is a marketer or an agency shopping for something to resell. I know because I am in those buying conversations, on both sides of them. That cluster is measuring supply as much as demand, and the industry is reading all of it as a market.
The damage is real. It is just not where the tools are pointed.
Further down in the same piece is the number I think about most:
AI Overviews cut clicks to the top-ranking result by 58%. Google’s referral share across their cohort fell from 35% to 24% in about a year.
So the loss is not happening because buyers left Google for ChatGPT. It is happening on Google, on results pages you already rank on, where the answer now sits above your listing and the click never happens.
You did not lose the position. You lost what the position used to be worth.
What that looks like on a real site
I have been inside a client’s Search Console for the last few weeks, an ecommerce catalog doing real volume. The numbers say the same thing from the other direction.
Over 28 days: 144,153 impressions, 843 clicks. A 0.58% click rate.
The top 24 non-brand queries produced 61,718 impressions and 92 clicks. That is a 0.15% click rate, at average positions between four and ten. Forty-three percent of the site’s impressions returning eleven percent of its clicks.
One term is worth sitting with. 28,238 impressions, 18 clicks, average position 6.9. That is 0.06%, roughly an order of magnitude below any published benchmark for that position.
And the one that ends the argument: a commercial-intent query at average position 2.6, with 757 impressions and zero clicks. Not a low click rate. Zero.
Position 2.6 with nothing to show for it is not a listing quality problem. You cannot write a better title tag out of that. Those impressions were counted, and they were never clickable.
I want to be careful here, because this is the point where the story gets convenient for me and I have not proved it. I cannot tell you from this data how much of that gap is AI Overviews specifically, versus other surfaces that push the blue links below the fold, versus listings that genuinely lacked the weight to be picked. Search Console does not separate them cleanly. What I can tell you is that the gap is far too large to pin on listing quality alone, and that everyone selling you a fix for it is guessing at the same split I am.
The defense is older and more boring than the category being sold
Same client, same data, split by branded and non-branded:
Branded queries: 1,082 impressions, 179 clicks, 16.5%
Non-branded queries: 143,071 impressions, 664 clicks, 0.46%
Branded search is 0.75% of what this site is shown for and 21% of what it earns. Over the trailing year it holds: 1.5% of impressions, 24% of clicks.
The single comparison: the brand name as a query returned 473 impressions and 124 clicks. The top category term returned 28,238 impressions and 18 clicks.
One brand query out-clicks twenty-eight thousand category impressions by seven to one.
That ratio is what actually survives an AI answer sitting on top of the results page. When someone searches your name, they are not asking to be told the answer. They already have the answer and they are navigating. No summary intercepts that, because there is nothing to summarize.
Every dollar of the AI visibility budget is chasing a channel at 0.32%. Meanwhile the thing that already works, that nobody is selling because it cannot be productized into a monthly subscription, is sitting at a sixteen percent click rate.
What I actually think is happening
Three things at once, and the industry is only pricing one of them.
Category search is being eaten. Broad informational and comparison queries are exactly what an AI summary answers well. Those impressions will keep getting counted and keep failing to convert. Treating them as recoverable opportunity is how a marketing budget gets spent twice.
AI referral traffic is real but early. 0.32% is not zero, and the growth curve matters more than the level. Getting cited by models is worth doing. It is not worth reorganizing your entire marketing operation around this year, and the tools measuring it are ahead of the thing they measure.
Branded and direct demand is the only layer that got more valuable. It was always the expensive, slow, unglamorous one. It is now also the only one an answer engine cannot sit on top of.
What we actually do about it
This is where I should tell you RevX sells AI visibility tracking. We do not, and the data above is why.
Three things get worked instead, in this order.
Make the site legible before anything else. Not for AI specifically. For everything. On a recent engagement the client’s product catalog was submitting 158 URLs for indexing and 74 of them were dead ends to an anonymous visitor, and the product pages published no price and no availability in machine-readable form at all. Eight days and four files later, pages eligible for shopping surfaces went from 0 to 66 and pages carrying review stars went from 0 to 295. That work makes you eligible on every surface at once, including the AI ones, and it costs a fraction of a monitoring subscription. It is also the only layer where the fix is unambiguous.
Stop paying for impressions that were never clickable. The audit above, run properly, usually retires a chunk of the keyword strategy. Terms sitting at good positions with sub-half-percent click rates are not underperforming content. They are answered queries. Reallocating that budget is worth more than any tool that tells you about it monthly.
Build the branded layer, which is the slow unglamorous one. Sixteen percent versus half a percent is the whole argument. Nobody productizes this because it cannot be delivered as a dashboard, and it takes quarters rather than weeks. It is also the only demand an answer engine cannot intercept, and everything above it gets cheaper once it exists.
If you want the first one run on your site, that is what RevX does and it is the cheapest engagement we offer, because it needs nothing from you beyond access. No content deadlines, no meetings. If you want the third one, that is a longer conversation and it is mostly not a technical problem.
The ten-minute version
Open Search Console. Filter to the last 28 days. Sort your queries by impressions and look at the top twenty.
If a term has thousands of impressions, a position under ten, and a click rate under half a percent, do not send a copywriter at it. Go look at what the results page actually shows for that query. If the answer is sitting above your listing, those impressions were never yours and the honest move is to stop counting them as opportunity.
Then split branded from non-branded and look at the two click rates side by side. That ratio tells you what your name is worth, in clicks, today.
Mine came back at seven to one. I would like to know what yours says. Reply and tell me, and if the number is ugly I will tell you which of the three above it points at.
Source: Ryan Law, “AI Marketing Trends 2026,” Ahrefs, September 16, 2026. https://ahrefs.com/blog/ai-marketing-trends/