AI positioning as a separate service: what you're really paying for, and when you're paying twice for the same work

Agencies already sell “positioning in ChatGPT”, “AI SEO” and GEO (Generative Engine Optimization) as a service with its own price list. Clients ask whether this is another fad with an invoice attached or real work. On this question I have data instead of hunches: on my sites, traffic from assistants grew eighteenfold in half a year, and on one of my platforms it converts 2.5x better than traffic from Google. The channel is real. The question is whether the service sold under that name is real to the same degree.

Take the service apart

Strip the naming layer off the offers and “AI positioning” consists of concrete activities. It pays to sort them by one criterion: is this work a good SEO agency should already be doing under the contract you may already be paying for.

Overlapping with classic SEO, which many clients already pay for: page speed and Core Web Vitals, correct semantic HTML, content available without executing scripts, JSON-LD structured data, content that answers real user questions instead of generalities, crawlability. An assistant hunting for sources behaves like a demanding crawler, so every one of these works for both channels at once. By my estimate this is roughly four-fifths of all “AI optimization”. I broke it down in detail in my guide to optimizing sites for assistants.

Genuinely new are the elements a typical SEO contract from two years ago did not include: a conscious robots.txt policy toward AI bots (GPTBot, ClaudeBot, PerplexityBot, Google-Extended), an llms.txt file, presence in the sources models build answers from (tool directories, industry roundups, directly quotable content), writing for long conversational questions instead of short phrases, and monitoring the effect in GA4 by session source. Since May 2026 there is also the Agentic Browsing audit in Lighthouse, which can measure part of this readiness.

That proportion leads to a plain conclusion. If someone proposes “AI positioning” as a separate retainer next to your existing SEO contract, then for most line items of that price list you are paying a second time for work that should already be done. The more honest construction, which part of the market itself uses, is extending the scope of the existing service with a GEO module, with the budget growing by the new activities, not by the whole thing again.

What nobody can honestly promise

The second problem with the offers lies in the promises. In classic SEO you can report a phrase’s position, because a search result is reasonably stable and verifiable. A language model’s answer is not: the same question asked twice can point to different sources, the result depends on country, conversation history and model version, and the providers publish no ranking. There is no such thing as “a position in ChatGPT” that can be guaranteed or honestly charted.

What can be measured honestly: visits from assistants in Analytics (for me these are specific numbers: 306 sessions and 15 conversions from chatgpt.com on a single platform), citations of your site in answers, spot-checked with a fixed set of questions, and the site’s technical readiness in an audit. An offer that promises “AI visibility” without saying which of these three measures the effect will be settled in is promising something unmeasurable.

How to recognize an honest offer

From client conversations and from reading the offers, I would compress the test to four questions.

One: does the opening audit show your site’s concrete gaps (bots blocked in robots.txt, missing structured data, content unavailable without JS), or does it only dramatize the market shift. Two: is the scope a list of activities you understand, or a set of names like “GEO strategy”. Three: will the effect be reported in your GA4 as sessions and conversions from AI sources, or as “increased visibility” with no definition. Four: given an existing SEO contract, is the new service an extension covering new activities, or a second price list for mostly the same ones.

Four good answers mean you are talking to someone who does the work. Two bad ones mean you are buying the channel’s name, not work on your site.

My answer to the title question

Does paying for AI visibility make sense? For the work that builds it, yes, because the channel grows faster than anything else in my statistics and brings users with concrete intent. For a separate service duplicating SEO, no, because eighty percent of that work is simply a well-made website.

For my clients this work is part of website care, not a separate line item: the same hands look after speed, structured data, content, and making sure assistant bots can get in and have something to read. If you already have an SEO agency, send them this post and ask which of the new activities are in your scope. A good one will answer with a list. That, too, is a test.