How to measure PR impact on AI visibility: the tools and method we use
A practitioner method for measuring whether press and earned media change what AI engines say about a brand, and the platform we run client programs on.
PR and AI visibility are now tied together in a way most comms teams have not caught up with. Third-party coverage is one of the main inputs the engines use when they decide which brands to recommend. So when a client asks "did that placement help?", they are asking a fair question. The honest answer takes a method, and it comes with limits we say out loud.
What we can and cannot say
We cannot prove that a single article caused a single change in a ChatGPT answer. Engines draw on many sources, answers shift between runs, and models change. What we can show is a chain of evidence that a client would accept in a board update:
- The placement exists and is readable by AI crawlers.
- The engines cite it for prompts that matter to the client.
- The client's visibility on those prompts moved over the same window.
- AI-referred visits to cited pages followed.
Each step is measurable. Together they make a defensible story. Any one of them alone does not.
Set the baseline before the campaign
Before outreach starts, we agree on a prompt set with the client: category questions, comparison questions, and a few brand questions. We run it for a few weeks to establish where the brand stands. Visibility Score, Share of Voice, Sentiment and Position in Answer give us the starting numbers. Without a baseline, every later chart is an argument.
Track the placements themselves
When a placement goes live, we add the URL to the Page Tracker in Promptwatch, the platform we run client programs on. The Page Tracker watches specific URLs, ours or a competitor's, so we can see whether the engines start citing the piece. Citation Trends adds the shape of the story: when the page was first cited, how fast the citation rate ramped up, where it peaked, and how long before it decays. That last point matters for PR. Some placements get cited for a month and fade. Others keep getting cited. It changes how much a client should pay for a similar one next time.
Offsite mentions covers the other half of PR coverage: third-party pages that name the brand without linking to it. Plenty of press coverage looks like that. Reddit, YouTube and LinkedIn have their own citation reports, which helps when a campaign includes community or video work.
Check the crawl side
An article can sit on a site that blocks AI bots. Agent Analytics logs AI crawler visits in real time, IP-verified, which tells us whether our own pages are being fetched by bots such as OAI-SearchBot, ClaudeBot or PerplexityBot. That matters for owned content that supports the PR push, such as a data study or a press page. If those pages are never fetched, the fix is technical and should happen before more outreach.
Compare periods and report
For the report, we use A/B period comparison on the dashboards: the four weeks before the campaign against a window after the coverage has had time to be crawled. Agent Chat can build the charts and a PDF report from live data, which saves us a day of formatting. Visitor analytics then adds AI-referred traffic and conversions per page, so the client sees whether any of the cited coverage sent people anywhere.
What the PR suites do and do not cover
Clients often already own a media monitoring suite. Meltwater and Cision both now sell AI visibility features inside their platforms, and they start from the media side. Neither publishes a crawler-log or conversion product that we have seen documented. We use them for coverage data and Promptwatch for the AI-side measurement. Pricing for those add-ons is not published, so check with the vendor.
Plan notes
Essential ($95 a month) covers the dashboards, citation reports and content gap analysis, but not crawler logs. Professional ($245) adds crawler logs, custom reports and Data Studio. Business ($579) adds the automated Content Agent, state and city tracking, and Ads Radar. The agency self-serve plans, Kick-off ($199), Growth ($399) and Scale ($799), include unlimited projects and crawler logs, which suits an agency measuring PR for several clients. A free Explore plan offers 10 prompts on ChatGPT only for a quick test.
The reporting line we use
"Coverage in these publications was followed by citations on these prompts and an increase in visibility on this topic. We cannot prove cause, and here is the trend line." Clients accept that sentence. They do not accept a claim of causation that falls apart when someone asks the second question.