Key takeaways
- Clients now ask "are we showing up in ChatGPT?" as often as they ask about Google rankings, and a screenshot is no longer an acceptable answer
- A credible report separates leading indicators (share of voice, citation rate) from lagging ones (AI referral traffic, conversions), and never mixes them without labeling which is which
- Fixed prompt panels matter: if the prompt set changes between reporting periods, the whole comparison is junk, no matter how good the dashboard looks
- Citation volatility is real. Reddit's share of ChatGPT citations dropped from roughly 3.8% to 0.5% in a single day in mid-August 2026, according to Promptwatch Data, so single-platform wins need a volatility caveat, not a victory lap
- GA4 undercounts AI referral traffic badly. Industry estimates put visible AI referrals at only 30-40% of the real number, so agencies need to say that out loud in the report, not hide it
Why the old reporting template doesn't work anymore
I've read a lot of agency reports in the last year that bolt "AI visibility" onto the bottom of a standard SEO deck, usually as a single slide with a screenshot of a ChatGPT answer and a caption that says something like "we're mentioned here." That's not a report. That's a trophy photo.
The problem is that AI visibility behaves nothing like a Google ranking. There's no stable position 1 through 10. The same prompt can return a different answer five minutes later, with different sources, different wording, and sometimes a different brand mentioned entirely. Clients who don't understand that will panic over normal noise, and clients who do understand it will ask pointed questions your screenshot can't answer, like