people started asking this as a monitoring question. It isn't one anymore.
The direct answer: run a fixed set of real buyer prompts through ChatGPT and Perplexity on a weekly schedule, log mentions and citations per engine (never averaged), confirm the Perplexity signal against GA4 referral data, and — the step almost every tool on the market skips — turn every gap you find into a published article before the next check. Tracking tells you where you're invisible. Only publishing fixes it.
Tracking AI Visibility Is Now a Revenue Question, Not a Vanity Metric
AI-referred visitors convert dramatically better than traditional organic search visitors. Independent analyses converge on the same range: a 12.3-million-visit dataset found AI search converting at 14.2% against Google organic's 2.8%, and Semrush's widely cited study put the multiple at 4.4x more valuable than organic search traffic by conversion rate. That gap exists because an AI-referred visitor has usually already been pre-qualified inside the chat conversation before they ever click through.
Despite that, most brands still aren't watching this channel. Only 22% of marketers currently track AI visibility, even as AI platforms drove 527% more referral traffic in just five months of 2025, per Previsible's AI Traffic Report. That's the gap worth exploiting: the brands building a tracking-and-publishing workflow now are compounding an advantage while three-quarters of the market flies blind.
Step 1: Build One Recurring Prompt Set, Not a One-Time Check
A single spot-check tells you almost nothing, because AI answers are not stable week to week. Citation churn is the norm, not the exception — one large-scale analysis found ChatGPT replaces 74% of its cited sources on a given prompt within a week, and only a small fraction of citations survive three identical runs of the same prompt. If your last check was a quarterly audit, it was already stale before you read the report.
The fix is a fixed, recurring prompt set — the actual questions a buyer would type right before choosing a tool in your category, not branded queries that already contain your name. Run the same prompts on the same cadence every time so movement is comparable week over week, not an artifact of asking a slightly different question. Weekly is the realistic floor for a fast-moving category; monthly is acceptable for slower ones, but quarterly checks miss the churn entirely. For the retrieval mechanics behind why this churn happens, see How AI Search Decides What to Cite.
Step 2: Score ChatGPT and Perplexity Separately — Never Blend Them Into One Number
This is where most tracking setups quietly mislead their own team. ChatGPT and Perplexity pull from different pools and reward different signals: ChatGPT leans on brands embedded in its training data, while Perplexity is grounded in live web results and rewards recency. When Mentionable ran the same prompts through both platforms, the top recommended brand matched only about 42% of the time overall — and dropped to 25–30% overlap in niche, less-consolidated categories.
That means "share of voice" is not one number. It's at least two, and they can move in opposite directions in the same month.
| Signal | ChatGPT | Perplexity |
|---|---|---|
| Primary basis | Training-data recall + trusted encyclopedic/blog sources | Live retrieval, weighted toward recency |
| Rewards | Established brand presence, structured explainer content | Fresh pages, high fact density, recent publish dates |
| Overlap with the other engine | ~25–42% brand match on the same prompts (Mentionable) | Same range, same source |
| GA4 referral trackability | Improving via GA4's native AI Assistant channel | Most reliably trackable AI source in GA4 — passes the perplexity.ai referrer cleanly |
If your dashboard reports a single blended "AI visibility score," you are averaging away the exact information you need to decide where to act next.
Step 3: Confirm the Signal in GA4 — Perplexity First
Monitoring tools tell you when you're mentioned; they can't tell you whether anyone clicked. That's a job for GA4. Of the major engines, Perplexity is the most reliably trackable in GA4 because it consistently passes the perplexity.ai referrer in browser sessions, surfacing in standard referral reports even without custom configuration. ChatGPT and Gemini traffic is messier and more prone to landing in "Direct" when referrer headers get stripped.
The practical workflow: if your prompt tracking shows rising Perplexity mentions for a target query and your GA4 perplexity.ai / referral line is also climbing for the matching landing page, you have two independent signals confirming the same trend — a monitoring tool alone can't give you that confirmation loop.
Step 4: Know Where Monitoring-Only Tools Stop Being Useful
The category leader in pure monitoring, Ahrefs Brand Radar, illustrates the ceiling of a tracking-only approach. A full review found it running $828–$1,148/month for full coverage on top of an existing Ahrefs plan, with no Claude or Grok coverage, no GA4 attribution, and no content production layer — it will tell you that you appeared in an AI answer, not what to publish next to fix the answers where you didn't.
That's the structural gap in this category: dashboards are common, closing the loop back into published content isn't. Aeolo's own workflow is built around that gap rather than around it — it finds blog topics directly from a brand's URL and visibility data, and keeps a weekly publishing cadence running so gaps don't sit untouched between checks.
Step 5: Close the Loop — Convert Every Gap Into a Published Article
Once your weekly prompt run produces a gap list, the highest-leverage move isn't a longer dashboard review — it's picking the highest-priority gap and writing the article that answers it, the same week you found it. A tracking report that sits in a spreadsheet for a month is a report about a version of AI search that no longer exists by the time anyone acts on it, given the source churn described in Step 1.
A workable loop looks like this:
- Run the fixed prompt set across ChatGPT and Perplexity.
- Log mentions and citation rates per engine — not blended.
- Rank the gaps by stage: comparison-intent queries first, foundational "what is X" queries last.
- Draft and publish one article against the top gap, with real inline citations and a structure AI engines can extract cleanly.
- Re-run the same prompt set the following week and check whether the published article shifted the mention rate.
- Repeat.
This is the step a pure monitoring subscription cannot do for you regardless of price, because publishing requires an editorial workflow, not another dashboard. It's also the reason a weekly cadence matters more than a bigger one-time report: each loop gives you a before/after comparison on the same prompts, so you can tell whether the article actually moved the needle instead of guessing.
If you're tracking gaps by hand today, Aeolo runs this loop end to end — surfacing the weekly gap list and turning the highest-priority one into a published article on the same cadence, instead of leaving the fix as a separate step.
FAQ
Is a monthly AI visibility check enough?
Monthly is workable for slow-moving categories, but treat it as a floor, not a target. Given that ChatGPT replaces the majority of its cited sources for a prompt within a week, a monthly cadence will miss meaningful swings in fast-moving or highly competitive categories.
Should I track ChatGPT and Perplexity with the same prompt set?
Yes — use the same prompts on both engines so the comparison is apples to apples, but score and report them separately. Averaging the two hides the fact that the two engines only agree on the top recommended brand roughly a quarter to two-fifths of the time, depending on how consolidated the category is.
Why does my GA4 show almost no ChatGPT traffic even though monitoring shows I'm mentioned?
A mention in an AI answer doesn't always produce a trackable click — some sessions lose referrer data and land in "Direct," and desktop/app clients pass referrers less consistently than browser sessions. Perplexity is the exception; it passes its referrer cleanly enough to show up in standard GA4 referral reports without extra setup. Treat GA4 as a confirmation signal for click-through, not a full picture of mention volume.
Is a dedicated monitoring tool like Ahrefs Brand Radar worth it on its own?
It depends on what you need next. A full review found the tool costing $828–$1,148/month for full coverage, without Claude/Grok tracking or GA4 attribution — useful if you already have a separate content pipeline to act on what it finds, but incomplete if you need the monitoring and the publishing loop to run together.
What's the fastest way to know if a published article actually improved my AI visibility?
Re-run the exact same prompt against the same engines the week after publishing, and compare the mention rate to the prior week's run on that same prompt. That controlled before/after comparison — same prompt, same engines, one week apart — is more reliable than watching an aggregate score drift, precisely because of how much weekly source churn already exists in these engines' citation patterns.
