On 12 November 2026, Duane Forrester published an analysis in Search Engine Journal arguing that the traffic pool is shrinking. He cites Kevin Indig’s Similarweb analysis of 10,000 domains each in the US and France, from August 2024 to August 2026. Organic visits fell 13.2% in the US and 24.4% in France. Paid search grew but replaced only about a third of the lost US organic visits. AI referrals were the fastest-growing channel, at about one visit in 200.
Forrester’s point is that comparing AI referrals with Google organic only tells you which stream is smaller. The better question is whether the visit still matters, or whether the transaction can happen without it.
The excitement is about which channel wins
Most of the conversation I see goes like this. Organic is shrinking, so where does the traffic come from next? Paid, email, AI referrals, social. Everyone picks a favourite and argues for it.
I would first ask whether that question is worth asking at all. Is there a real need to find the next channel? Only if the visit is still the thing that produces the sale. For a traffic-dependent business, the visit was always a step toward the sale. A visitor arrived, read, compared and decided.
That is the part that changed. Forrester cites a McGill study where agents were given 100 randomised hotel listings across 5,000 sessions on four LLMs. Position had weak effects. Every model converged on the same listing, which took 78.2% of bookings. The authors conclude that the attributes shown matter more than placement.
So the agent did the comparing and the deciding. No visitor read your page.
What changed and what didn’t
What changed is who does the evaluating. An agent reads your information, weighs it against your competitors’ and picks one.
What didn’t change is that someone still has to buy from somebody. Forrester is also careful about timing. Google’s Universal Cart launched in the US on 15 October, but a May scan found only 26 sites publishing their own UCP profile among more than 3 million. OpenAI pulled back Instant Checkout in March 2026 and moved purchases back to merchant sites. Much of this is early.
My view on agentic commerce overall is that it is too early to call. Big platforms may block agents or welcome them, and the giants can afford to wait and see. Smaller businesses can’t. For them the shift is already decided, and the move is to get data-centric.
What a small operator should do
Take a small clinic, a shop or a service business. The owner’s instinct is to ask whether to spend on ads, email or “AI SEO” to get the visits back. That treats the visit as the goal.
I would spend that effort on the offer instead. Your information will be read in many places, not only on your website. Wherever an agent reads it, the offer has to be precise. That means specific, measurable attributes that differ from your competitors’, such as price, rating, delivery terms, packages and areas served. If your page says “affordable, trusted, friendly service”, an agent has nothing to compare. If it says what you charge, how many reviews you have and how fast you deliver, it does.
This is the same reason I say ranking alone does nothing. On stridec.com, non-branded positions 1 to 3 got a 0.08% click-through rate between June and August 2026. A page-one position was only ever a step toward the sale, and so was the click.
Forrester’s list is practical on this. He suggests picking the three to five attributes agents can read, then closing gaps or stating trade-offs clearly. He also says to publish policies such as returns and cancellation as their own fields. Service businesses should publish packages, prices, areas and availability in machine-readable form. And your name, pricing, hours and specs should match across your site, profiles and directories.
Everyone also asks their agent the same question slightly differently. To catch that, your information has to cover a much broader spread of questions, with varied and detailed attributes.
Where I do this myself
AeroChat publishes many pages that each target an integration with a specific piece of software. Each page catches a long-tail question. I do it because agents and AI search engines can only quote what is written down in specific terms.
I would not read this as a case for dropping the other channels. Forrester notes email grew about 16% in the US sample, the steadiest gain of any channel. Keep what works. I just wouldn’t spend the budget trying to guess which channel refills the pool.
What to do this month
- List the three to five attributes a buyer or agent would use to compare you. Write each one as a number or a plain fact.
- Check where you are the same as your competitors. Either change the offer or say clearly what the trade-off is.
- Put your prices, policies, packages and availability on your pages as plain, specific text, and keep them consistent everywhere your business is listed.
- Track your total inbound sessions by source each month, as Forrester suggests, so you can see the pool itself and not just one channel.