On 5 November 2026 Search Engine Journal published a sponsored piece by Bartek Jazwinski, Director of Product Management at WebPros, titled “AI Visibility vs. Search Rankings: Is Your Agency Measuring the Wrong Thing For 2027?” It argues that agencies still leading their reports with sessions and rankings “are on borrowed time”. In their place it proposes three gates for measuring AI visibility: inclusion (is the brand named or cited in the answer), framing (is the description accurate) and endorsement (is the brand recommended for the buyer’s specific need).

The claim, and what is actually new

The line I agree with most is this one: “A ranking is a position in a list. The user still has to scan the options and choose one. An AI answer isn’t a list, it’s a verdict.”

That is the whole shift in two sentences. So what is new here, and what is not?

What is not new is the idea that rankings no longer describe the outcome. Ranking is dead, and my own data says why. On stridec.com, non-branded positions one to three on Google earned a 0.08% click-through rate between June and August 2026. Page one, top three, almost no clicks. A ranking report that showed those positions would have looked like a win. The finish line had moved.

What is new is that a mainstream trade publication is telling agencies to rebuild the report around whether the answer names you. The article also makes a point I would underline. It warns that “citation share is turning into the new keyword ranking, and it carries the same bad habit”: counting mentions without checking what the buyer was asking for. That is right. Counting citations the way we used to count positions swaps one vanity number for another.

One thing the article cites cuts the other way. Ahrefs panel data puts AI assistants at under 1% of site traffic in September 2026. Someone will read that and say the whole thing can wait. I read it differently. The click was already collapsing at the top of Google before AI assistants sent any traffic. Measuring by traffic is what hid that. The answer is the surface now, and the question is whether you are in it.

Three gates is right. One number goes first.

Inclusion, framing and endorsement is a sensible ladder. For an agency building a measurement practice, it is the right direction. But an operator reading a report does not need a ladder. They need the first line to tell them whether things are improving, stagnant or dropping.

So I collapse it. The number at the top of the report is how often the answer names your brand, across a fixed set of the questions your buyers actually ask, on a stated date. Last month’s number sits next to it. Everything else supports that line.

Citations are the leading indicator underneath it. If the engines are using your pages as a source, naming usually follows. If they are not, naming is unlikely. So citations go in the report, but second, as the thing that predicts next month’s top line.

Rankings are the relic. Keep them in an appendix if a client still asks. They tell you whether the groundwork is in place, and nothing about the outcome.

The two can break separately

Here is why the top number has to be naming and not position. In August 2026 AeroChat, my own product, ranked third on Google for “best Shopify chatbot” and was not named in the AI Overview for the same question. A ranking report would have called that healthy. When I went through the full list of sources behind that answer a month later, one article named AeroChat first and none of the others mentioned us at all, so the AI didn’t either.

Ranking and being named are separate. Find out which one is broken before you fix anything. A report that leads with position cannot tell you that. A report that leads with naming, with the citations beneath it, points straight at the cause.

What this looks like for a clinic or a shop

Say you run a dental clinic and your agency sends a monthly report. The 2027 version should open like this. Twenty questions a patient actually asks before booking, the same twenty every month. The share of answers across Google’s AI Overview, ChatGPT and Gemini that name your clinic, dated. The same share last month. Then, below it, which of your pages and which other people’s pages the answers drew on.

AI answers are regenerated on every search and change week to week, so the date is part of the number. A naming share without a date cannot be compared with next month’s.

What to do

Ask your agency for one number at the top of the next report: how often the answer names you, on which questions, on which date. If they cannot produce it, that is the finding. Keep citations as the second line and watch whether they move before naming does. Move rankings and sessions to the back. Then read the first line each month and decide from that.

Source: Search Engine Journal: AI Visibility vs. Search Rankings: Is Your Agency Measuring the Wrong Thing For 2027?