A monthly report can show first-page rankings and stable organic traffic, while reporting fewer qualified leads. In a sponsored op-ed by WebPros, whose opinions are those of the sponsoring publisher, the author argues that agencies must measure something other than ranking: visibility in AI tool responses, which he proposes evaluating in three steps.
An AI Response Is Not a List
A search engine ranking gives a position in a list, and the user must still browse the options and choose one. A response generated by an AI model works differently: it resembles a verdict, as the model has already compared the options. When the user reads the response, the selection may already be made, meaning a brand sometimes never had a chance to get the click.
The op-ed distinguishes between citation, mention, and recommendation. A model can cite a company among five options while regularly recommending a competitor. Conversely, a brand can appear in a response without being presented as a suitable solution. The text argues that the frequency of mentions, taken in isolation, is a misleading indicator.
Three Steps to Measure Visibility
The first step, inclusion, is about whether the brand appears in the response. To measure it, one must ask a defined set of questions related to the client’s business, test them on the monitored platforms, then report the number of responses mentioning the brand out of all tested responses. The source gives the example of a brand present in 12 out of 40 responses, an observed rate of 30%. It specifies that this figure informs about the tested questions, not about all AI searches.
The second step concerns framing: is the description of the company accurate? A technically correct citation can present a brand as a cheap option or an outdated player, which can harm more than absence of citation. The source cites the case of a B2B software company whose old product pages described the tool for freelancers, while it now targets large enterprises.
The third step, endorsement, checks if the brand is explicitly recommended on purchase questions. The text describes the case of an AI that cites a client’s product, correctly describes its features, then recommends a competitor it deems better suited for small teams. Inclusion and framing are successful, but endorsement fails. The three steps are distinct measures, not a conversion funnel: an increase in citations accompanied by a drop in the share of recommendations can therefore be a warning signal.
Linking Figures to Business Results
The source urges caution in interpreting links to business results. It recommends tracking separately the progression of brand searches, identifiable visits from AI, and leads or sales linked to these channels. Brand searches also increase with press relations, paid campaigns, and seasonality.
A Similarweb study on post-ChatGPT recommendation behavior reports that, within seven days, users who saw a brand recommended were more likely to visit it than its competitor. When ChatGPT recommended American Express, 7.2% of users later visited its site, versus 3.1% for Capital One. Conversely, 14.2% visited Capital One after its recommendation, versus 3.8% for American Express. The study is based on a US desktop panel with voluntary opt-in, covering July to December 2025, and on consumer brands. Its results therefore do not necessarily apply to small businesses or B2B.
According to the Ahrefs Web Analytics opt-in panel, AI assistants accounted for less than 1% of website traffic in September 2026. The text specifies that this figure measures traffic received by participating sites, not the share of AI tool usage that generates an outbound click.
The client report proposed in the source links each finding to an interpretation and an action. Its figures, presented as illustrative and not actual results, show mentions rising from 30% to 45% of tested responses, and accurate descriptions from 70% to 85%. Explicit recommendations rise from 15% to 25% of tested purchase responses, and clicks from brand searches increase by 12%, without this rise being attributed to AI.
