Skip to main content
x

AI search visibility needs better measurement than brand mentions

AI search visibility needs better measurement than brand mentions

(Bibek Thapa) - By Bibek Thapa

Edited by Jenn Moon, Editor at Featured News · Nov 17, 2026

Traditionally, SEO has been measured through a variety of metrics, including rankings, impressions, clicks, click-through rate and organic sessions. This will help you understand if your content is being seen by your target audience.

However, with the introduction of AI in search, we’re seeing additional outcomes to track. Your content could be mentioned as a source in AI answers, your brand could be mentioned without any clickable citations and your website could gain visibility without any sessions.

This can cause issues when measuring your SEO. However, Google has started to bridge the gap by adding a new report called the Generative AI performance report in their Search Console. Google completed the report's worldwide rollout on August 31, 2026, although sites without enough impressions may not see it. The report displays how many impressions your website has had within supported generative AI features in Google Search (including AI Overviews and AI Mode).

Website owners no longer have to rely on manual methods alone to understand how they’re doing in Google’s AI search. However, this is only for Google’s search engine. It will not show you how many citations your website has in ChatGPT, Gemini, Perplexity and other AI platforms. This means SEO professionals are having to piece together different ways of measuring how they’re doing from multiple partial sources.

Mentioned vs. cited

The main thing to understand is the difference between a mention and a citation.

A mention is when your website, company, product or brand is mentioned in the AI answer. A citation is when your website or page is cited as a source.

These don’t always go hand in hand. Some websites will be mentioned but not linked to. Other websites will be cited as a source, but your brand won’t be mentioned much. If you’re counting mentions as citations, your website will appear to be a more important source of information than it actually is.

When looking at how your website is performing in ChatGPT, Gemini, Perplexity and Google AI Mode, each platform will cite sources differently. A better way to measure this is keeping track of:

  • Brand mentions

  • Citations

  • Exact URLs cited

  • Repeated citations

  • Visible referrals

Each of these will answer different questions but all together will give you a much clearer idea of how your website is performing in AI search.

Why it’s important to run multiple tests

It’s not useful to just run one test. There are multiple variables in the answers from each platform, and even the same question can be answered differently. The pages cited can also change.

I ran the same prompt on ChatGPT, Gemini, Perplexity and Google AI Mode in September 2026:

"Search the web for Anobee and cite the sources you use."

It mentions the brand, which will influence the results. It was created to understand how each platform found and cited pages from my website. It does not show how well the AI platforms find your brand naturally and isn’t representative of other types of queries.

The test wasn’t the best either. I was running them from Nepal, using a logged in ChatGPT Plus and Gemini Flash account. Perplexity was run with no login.

In a September 18 test, ChatGPT cited six pages from my website, and Gemini cited one page. Perplexity had 15 sources cited, 11 of which were from my website. Google AI Mode had no citations.

Looking at the URLs, 13 different pages were cited on September 9, 11 on September 14, and 14 on September 18. This means I had a total of 25 unique URLs cited, but only four appeared on all three testing dates.

This shows that although my website was cited on all three testing dates, most of the pages cited were different. If we were just looking at domain level, we wouldn’t know this at all.

This is just testing using my own website and doesn’t prove anything about how the AI systems decide on sources.

You can miss out on URL changes

Citing at domain level will show you if your website was cited, but citing at URL level will show which pages were cited. Both should be tracked.

Using an example, suppose you have a website which writes about technical SEO, content optimization, analytics and digital marketing. One AI answer cites your technical SEO guide and another cites your analytics article. This would mean your website was cited twice at domain level. However, you would have two pages cited at URL level.

From my tests in September, I had 25 URLs cited but only four of them were consistent. If I was just tracking this at domain level, I would be exaggerating how often any specific page was being cited.

Another example of this would be from testing Perplexity on October 14, 2026. My homepage was cited but none of my other pages (About, Contact and Author). This doesn’t explain why I didn’t get citations on these pages but clearly shows why it’s more useful to track the exact URL that was cited.

Tracking at page level will show you what content is being cited regularly, what pages are being cited occasionally and what parts of your website aren’t being cited at all. It will also show you what was cited based on your specific criteria. It doesn’t show you how the platform ranks or decides to cite.

Creating a more useful way of tracking AI visibility

Although this doesn’t need to be complicated, having consistent testing, good tracking and clear definitions will be more useful than having a single visibility metric using an unreliable method.

Set out what you want to know

Start by asking questions your target audience would actually ask about your topic. Don’t create prompts to get mentions of your brand.

If you want to know how well your brand is being found, this is a separate goal. Asking for your brand in your prompt is different to asking a general question.

Keep track of your exact prompts

For each test, note down what platform you’ve used, the date, time, prompt and any variables that could impact the outcome of the test (logged in vs. not, available options, location, etc.).

Keep the results separate for each platform. If you don’t do this and just have a generic list, you’ll be drawing misleading conclusions. Having them separated will also help you see what has changed between tests.

Track mentions and citations

Note down if your website was mentioned, if there was a clickable source and if it was from your website.

If your brand was just mentioned and not linked to, this isn’t a citation. By separating the two, your report will be honest about how much visibility you actually had.

Track individual URLs

Note down the full URL of your cited pages. While knowing your website was cited at domain level is useful, it doesn’t tell you what content was cited.

Having this will show you if only a few pages are being cited or if it’s spread across your website. This will also help when comparing results from multiple tests.

Run multiple tests

Ask the same questions using the same criteria as above. Compare results from different dates and platforms. Carefully note down what wasn’t cited.

Don’t automatically assume that if your content wasn’t cited the same way, your content has improved or deteriorated. Multiple observations can highlight trends but won’t tell you why the platform responded differently.

Compare with your website analytics

Citations and referrals are separate. Just because a page was cited, doesn’t mean it gained any measurable referrals. Just because you had a referral from an AI doesn’t mean it was from a source that was cited.

GA4 has an AI Assistant channel in your default channel group. Google states some examples of sources for this are ChatGPT, Gemini, DeepSeek, Copilot and Grok. However, Google AI Overviews and AI Mode are not included. This traffic will be classed under Organic Search. Any traffic from Google’s own AI features will not show in the AI Assistant channel.

Within my GA4 data, I had 45 sessions in the AI Assistant channel between September 15 and October 14, 2026. 41 of these were from chatgpt.com.

I also had one link with the utm_source=openai parameter but no utm_medium. Traffic from that link appeared under Unassigned rather than AI Assistant in my GA4 report. This shows how tracking setup can affect the AI referral traffic reported in GA4.

Limitations

There are a few limitations to testing how visible your brand is in AI:

  • Different platforms. Just because you received a citation on one platform doesn’t mean you’ll get it on another.

  • Small sample. You’re only using a few prompts, so you can’t represent all queries. Don’t claim your small test represents the whole industry.

  • Answers change. Pages can change between tests.

  • Not all referrals. Someone could find your brand using an AI answer but not come back via a referral.

  • Limited reports. The Search Console report only shows visibility within AI Overviews and AI Mode. The GA4 AI Assistant channel shows visibility from other assistants.

While this doesn’t make the test useless, it means you need to be aware of what you’re measuring.

What this means for the future of SEO reporting

Traditional SEO metrics are still relevant in AI search. Rankings, impressions, clicks, organic sessions and engagement are still important. Google has shown they care about site owners being able to see how visible their website is in their new Search Console report.

However, you need to separate traditional search performance from AI visibility. The Search Console will show you how many impressions you had in AI Overviews and AI Mode. GA4 will show you how many sessions you had from AI assistants. Manual testing will help you see how many mentions, citations and individual URLs you had across different platforms. These methods complement one another and cannot be swapped.

A good report will show you where the numbers came from, what they represent, what platforms were tested and what limitations there are. Don’t combine everything into a single AI visibility metric.

Final thoughts

There are four separate outcomes to track: brand mentions, citations, repeated URLs and measurable referrals. If you combine them all together, you’re hiding most of what is actually going on.

Testing multiple times and looking at my own analytics has shown me that while a domain might be consistent, the pages cited can be different. You could also be gaining impressions/referrals from AI that aren’t captured in any single report.

As an SEO professional, the next step is to improve how you collect, track and interpret the data you can see. Don’t try to push your results too far.

About the author

Bibek Thapa

Bibek Thapa is the founder of Anobee, an independent digital growth publication. His work focuses on SEO, AI search optimization, AI workflows, content strategy, website growth and productivity systems. He uses hands-on website audits, Google Search Console and GA4 data, and AI citation tests to explore how websites perform across search and AI-driven discovery.

This story was produced by Featured and republished with permission. Experts featured here are members of Connectively, Featured's expert network.

Article Link