Your AI visibility score is now a commodity, and moving it is still nobody's job

In two days, three vendors and one agency published versions of the same admission: the AI visibility score has become a commodity, and none of them can sell you the thing that moves it. Ahrefs, HubSpot and Semrush all pushed out material aimed at marketers already tracking how often ChatGPT, Gemini and Perplexity name their brand, and all three land in the same place. Counting is solved. Changing the count is unassigned work sitting between three teams that do not share a calendar.

The framing is bluntest in the Ahrefs webinar page carried on Search Engine Journal, which argues the measurement stage is ending. Queries about tracking AI visibility, the company says, far outnumber queries about improving it. In its words, "Nearly everyone can see where AI cites their brand; very few have a process for changing it." That is a vendor describing its own market, on a page bylined by SEJ's VP of sales and marketing, not a reporter — discount it accordingly, then check it against your own last quarter.

What the sellers are conceding

HubSpot's audit walkthrough, also on SEJ and labelled as sponsored by HubSpot, is the most useful of the set: it splits two numbers most dashboards mash together. A mention is your brand appearing anywhere in an answer. A recommendation is your brand put forward as an option the buyer should consider. The guide notes that a brand can be mentioned in most answers and recommended in almost none, and that recommendation is the figure tracking with pipeline. A single "AI visibility" percentage on your monthly report tells you neither.

The method underneath is deliberately unremarkable: take the commercial queries you already rank for, rewrite each as a question a buyer would type, run it a few times per engine signed out, and log which brands and domains come back. No software is needed for that. What software buys you is the schedule.

Most of the score sits on domains you do not own

There is a structural reason the number resists a dashboard. HubSpot's guide sorts the sources feeding AI answers into six types: review directories such as G2 and Capterra; Reddit threads and forums; third-party "best software" roundups; news coverage; reference pages like Wikipedia; and your own website. It puts your own site last, as usually the smallest slice even though it is the only one you control outright. Most of those categories are edited by somebody else.

So the most operational piece of the week reads like an org chart rather than a tool. Writing on MarTech, Zen Media founder Shama Hyder describes rebuilding her agency so PR, SEO and content all start from one list of the prompts buyers are typing, on a weekly loop rather than a quarterly review: read what shifted, place coverage and rebuild pages, then log what each prompt returned and in what order. Monitoring, in her account, is the logging, not a dashboard someone glances at. MarTech is owned by Semrush and carries that disclosure.

She cites SparkToro's read of Similarweb panel data for the first four months of 2026, putting 68% of US Google searches as ending without a click, and Pew Research finding users click a traditional result in 8% of visits where an AI summary appears, against 15% where none does. The answer is now the surface you optimise, and it is rebuilt from scratch every time someone asks.

The score is cheap. The thing that moves it is a standing weekly meeting between three teams that have never shared one.

What a software marketer should do differently

Semrush's own contribution, a post it wrote and Search Engine Land published to promote its Spotlight conference in London on 13 October, is the most nakedly commercial of the four and carries one figure worth keeping: of 1,094 categories it analysed, the company says only 15.2% had a clear owner in AI answers. Discount a vendor's own study as you like, but the direction is plausible for B2B software. Leigh McKenzie, its director of online visibility, describes the fix as a short list of things "done consistently week after week" — a process claim, from the company selling the product.

The reporting line and the working line have come apart. The working line is four habits nobody currently owns:

• Split the single visibility number into a mention rate and a recommendation rate. Budget conversations should hang on the second.
• Pull the source mix behind your prompts before commissioning content. If a directory and three roundups are doing the citing, a blog post is the wrong first move.
• Give the weekly loop a named owner. If PR, SEO and content read the same prompt list on a Monday, most of the coordination cost disappears.
• Audit directory profiles like landing pages: current pricing, correct categories, a steady arrival of reviews rather than one ageing batch.

Worth saying out loud, since this week's publishers will not: every substantial piece of AI visibility guidance in the trade press across these two days was written by a vendor, sponsored by a vendor, or published by an outlet a vendor owns. That does not make the advice wrong. It does mean the category has no disinterested benchmark and no agreed definition of a mention — so the number on your dashboard is a vendor's opinion with a decimal point on it.