Until now, visibility in AI assistants has largely been a black box. Companies could suspect that they appear in ChatGPT answers, but could hardly prove how often that happens and which information is used in the process. Hints of a beta version of a ChatGPT tool now suggest that this could change.

According to the available information, the tool is intended to enable two things: first, tracking impressions – i.e. how often a company appears in answers – and clicks that originate from within ChatGPT. Second, companies are to be able to store information about themselves and thereby influence how they are perceived in the AI service.

The classification is important: this is a beta version and hints, not a fully rolled-out product. Concrete dates, prices or the final scope of functions are not known. Anyone planning budgets now is doing so on suspicion.

Why this is relevant for website operators

In classic search engine marketing, the measurement chain has been established for years: a page ranks, generates impressions, is clicked, and the visit lands in web analytics. With AI assistants, this first link has so far been missing. Users ask a question, receive a summarised answer – and whether and how a company appeared in it remained invisible.

If impressions and clicks become available first-hand in future, a gut feeling becomes a metric. This also makes the optimisation of content for AI answers verifiable: measures can be compared with a development in numbers instead of only with samples of individual prompts.

At the same time, the goal shifts. In classic search, the click to your own website is the success. In an assistant, the answer can already deliver the information without anyone visiting the page. The mention itself then becomes a value in its own right – similar to reach in other channels, only at a point where concrete purchase decisions are prepared.

Self-disclosure as the second half of the story

The second part of the indicated tool is at least as interesting from a company perspective as the statistics: the option to provide information about your own company yourself. This is reminiscent of familiar patterns from other platforms, where maintained company profiles influence how a company is presented in search and maps.

In practice, this means: a further data source may emerge that has to be maintained – with all the familiar follow-up questions. Who is responsible internally? How do details on services, locations and contacts remain consistent with your own website, with industry directories and with press texts? Contradictory information across different channels is already a quality problem today; in AI answers it becomes more visible because systems summarise information instead of merely linking to it.

What you can sensibly prepare now

Regardless of when and in what form such a tool becomes available, preparatory work is worthwhile and pays off anyway:

  • Clean up the factual basis: Keep service descriptions, location data, responsibilities and company details on your own website complete and free of contradictions.
  • Use structured data: Machine-readable markup in the source code helps systems assign content correctly – this benefits both search and AI answers.
  • Think in questions rather than keywords: Content that answers concrete questions is easier to quote than pure product prose.
  • Establish your own baseline: Start documenting today, on a sample basis, how assistants describe your company. Without a starting value, there will be no comparison later.
  • Prepare analytics: Capture referrals from AI services cleanly in web analytics so that visits from these channels do not get lost in direct traffic.

Stay realistic

A beta version says little about how reliable the later figures will be. It remains open, for example, how impressions are precisely defined, at what granularity data will be available and whether access will be tied to conditions. The question of how strongly self-disclosed information actually flows into answers also cannot be answered at present.

A sober approach is therefore sensible: observe the development, improve your own content, but do not base a strategy on a product whose details are not yet fixed. Anyone who has robust content and clean data maintenance today will benefit regardless of which measurement tool ultimately prevails.

Classification for web and software projects

For web projects, this above all means: content structure and data quality become more important than individual optimisation tricks, because AI systems summarise information instead of merely listing pages. In a content management system such as TYPO3, central company details should be maintained in one place and distributed to all output channels instead of being duplicated multiple times in the page tree. And in project planning, it is advisable to provide for AI visibility as a separate measurement point in reporting – even if the data sources for it are only just emerging.

Sources