AI Visibility Is Becoming Measurable. The Numbers Still Need Interpreting

For a long time, one of the frustrations with AI visibility was simple: you could see that ChatGPT, Copilot and other answer engines were citing websites, but measuring what was happening to your own site was considerably harder.

That is beginning to change.

In February, Microsoft introduced AI Performance in Bing Webmaster Tools. It reports how often a website is cited across Microsoft Copilot, AI-generated Bing summaries and selected partner integrations, including cited URLs, citation trends and some of the phrases used when retrieving the content. (Bing Blogs⁠)

OpenAI also allows publishers to identify referral traffic from ChatGPT because links from ChatGPT search automatically include utm_source=chatgpt.com. (OpenAI Help Center⁠)

AI visibility is starting to leave a trail.

For businesses trying to work out whether they are actually being discovered through AI, that is progress. It also creates a new problem: having AI data is not the same as understanding what needs fixing.

A citation is useful. It is not a verdict on your website

Bing’s reporting is particularly useful because Microsoft is careful about what the figures mean.

Its Total Citations figure counts appearances as sources in AI-generated answers. Average Cited Pages tracks the average number of unique pages from the site cited each day. Page-level data shows which URLs are being referenced.

Microsoft explicitly says those figures do not indicate ranking, authority, performance or the importance of a particular page. (Search - Microsoft Bing⁠)

That is an important restraint.

A site receiving citations may still have pages with weak indexing, confusing structure or poor entity signals. Another site may have strong technical and content foundations but little measurable AI referral traffic yet.

The useful question is therefore not simply, “How many AI citations did we get?”

It is: why are some parts of the website discoverable while others are not?

Google adds another piece of the picture

Google has also expanded the visibility available to website owners. Search Console now includes dedicated generative-AI performance reporting covering features such as AI Overviews and AI Mode, while that activity continues to contribute to overall Search performance reporting. Google says the dedicated reports were rolled out worldwide by 31 August 2026. (Google Developers⁠)

So website owners can increasingly assemble evidence from several directions: conventional search performance, Google’s AI-influenced search results, Microsoft AI citations and identifiable ChatGPT referral traffic.

But those platforms are measuring different things.

Putting the numbers into one spreadsheet will not tell you whether an important page is difficult to crawl, whether its canonical signals are confused, whether the content clearly identifies the product or organisation being discussed, or whether an AI system can confidently understand what the page is actually about.

That is the layer IndexStream⁠ is designed to investigate.

Measure the causes as well as the outcome

IndexStream analyses the website foundations that support search and AI discovery, including indexing readiness, technical SEO, content, schema, entity signals and AI visibility. It turns findings into prioritised actions rather than simply producing another collection of numbers. (IndexStream⁠)

Rather than presenting citation counts as a verdict on a website, IndexStream looks for practical weaknesses in the site itself — the things a business can actually inspect, understand and improve.

That becomes increasingly useful as AI reporting develops.

If Bing tells you a page is being cited, that is evidence worth having.

If analytics tells you somebody arrived through ChatGPT, that is evidence worth having too.

IndexStream tackles the question underneath those outcomes: what can we see on the website itself that may be helping or obstructing its discoverability?

The objective is not to manufacture an “AI score” and pretend it predicts what ChatGPT will say tomorrow. It is to identify weaknesses that could be making a business harder for search engines and AI systems to understand and discover.

AI discovery is becoming more measurable. The businesses that get the most from that data will be the ones that use it to investigate why their websites are — or are not — being found.

Analyse your website with IndexStream and see what is worth fixing first.⁠