How to measure visibility when nobody clicks

Clicks were always a proxy. For years the proxy was good enough: if you ranked, people clicked, and the click graph moved with your visibility. That link has weakened. A searcher who asks for opening hours, a price range or the difference between two options often gets the answer above the results and moves on.

That does not make the exposure worthless. The person saw your brand as the source, may search for you by name later, or may type your web address directly the next time they need you. None of those later actions is credited to the original search in a standard report, so the click graph understates what SEO is doing.

The fix is not to stop counting clicks. It is to stop counting only clicks, and to build a small scorecard that captures exposure, demand and outcomes separately.

Impressions in the Search Console Performance report count how often a page from your site appeared in results someone saw. Google includes appearances in AI features such as AI Overviews within the Web search data but does not break them out, so impressions are a combined exposure number rather than an AI-specific one.

Ownership you have to measure yourself. Pick thirty to a hundred queries that matter commercially, check them on a fixed schedule, and record whether you hold the featured snippet, appear in People Also Ask, or are cited in the AI Overview. A rank-tracking tool can automate part of this; a spreadsheet is fine to start.

Branded search and direct traffic are the downstream signals. If people who met you in an answer later search your name or type your URL, those numbers rise. They are noisy — offline marketing, word of mouth and seasonality move them too — so read them as a trend, never as proof on their own.

In the Performance report, add a query filter that excludes your brand name to see non-branded demand. The filter accepts regular expressions, so one pattern can catch your brand and its common misspellings — useful for names people type in several ways. Then build a second view that includes only branded terms.

Compare matched periods — this quarter against the same quarter last year — so seasonality does not fool you. Look for the pattern: impressions up, position stable or better, clicks flat or down. That is the signature of answers being consumed on the results page, not of a ranking problem.

Then go page by page. Pages that answer simple factual questions usually lose the most clicks to answers; pages that help someone compare, decide or buy usually hold up better. That split tells you where to keep investing in answer-shaped content and where to give people a reason to click that a summary cannot.

• Filter: queries not matching your brand (regex catches misspellings)
• Compare: date range against the same period last year
• Columns: impressions, clicks, CTR and average position together
• Group: pages by intent — fact, compare, buy

There is no Search Console for ChatGPT, Gemini or Perplexity. The practical method is a prompt panel: twenty to fifty questions your buyers actually ask, such as ‘best CA firm in Indore for startups’ or ‘accounting software for UK sole traders’, run monthly and logged as named, cited with a link, or absent.

Answers vary between runs, users and dates, so treat a single check as an anecdote. Run each prompt more than once, note the date and the model, and judge the trend over several months rather than reacting to one week’s screenshot.

In GA4, build a custom channel group or exploration that captures sessions where the source is an AI assistant domain, such as chatgpt.com, perplexity.ai or gemini.google.com. These sessions are usually few, but they often come from people who have already asked for a shortlist, so check their key event rate as well as their volume.

No attribution software will fully credit an answer the buyer read without clicking. Self-reported attribution fills part of the gap. A simple dropdown or free-text field — Google search, AI assistant, social, referral, other — often surfaces channels the analytics never saw, such as ‘I asked ChatGPT’ or ‘I found you on Google Maps’.

In GA4, mark the actions that matter as key events: form submissions, click-to-call, WhatsApp clicks, booking confirmations. Then you can see whether branded and direct sessions — the ones zero-click exposure feeds — convert at a healthy rate.

Finally, put everything on one monthly page. When impressions, ownership and branded demand rise and enquiries hold or grow while clicks fall, you have a credible case that visibility is working. When all of them fall together, you have a real problem to fix.

A founder who sees organic clicks fall will reasonably ask whether SEO is broken. Answer before it is asked: open the report with leads and sales by source, then show impressions, answer ownership and branded search as the leading indicators behind them.

Annotate the timeline. Note when Google expanded AI features in your market, when you published or refreshed key pages, and when offline campaigns ran. Context turns a confusing chart into a story someone can act on.

• Page 1: leads and revenue by source, against last year
• Page 2: impressions, answer ownership and branded search trends
• Page 3: what changed, what we did, what we do next
• Appendix: clicks and CTR, with the zero-click context