Most of the work is a manual effort coupled with a mishmash of ideas, concepts, metrics, and nascent tools; there’s no Radian6 for AI visibility yet
When social media really took off somewhere around 2008-2010, it took a while for the PR software monitoring to catch up.
Some tools were one-offs that could measure corners of the social web. Others were taking data feeds from a provider, like Moreover, but it would take 24 hours to appear in your dashboard, which didn’t help with real-time information. Speed was important and most tools didn’t offer it.
Then a company called Radian6 came along. It was a media darling. You could basically search across social media with Radian6 the same way people search the web for information. They even opted for the vaunted “firehose” from what was then Twitter – an integration that provided granularity and depth to reporting you could not get from merely scraping the site.
The success that the company had prompted a whole range of PR software companies to invest in developing better ways to monitor social media, along with traditional news.
I think something like that is happening with AI visibility right now. We’re in the early stages. PR is just starting to think about what AI means for PR and marketing communications.
Measuring it is hard. There are no frameworks. Standardized metrics do not exist. It’s a complicated undertaking to get data (if that’s even possible at the moment) from the various LLMs to provide a comprehensive view.
So, against that backdrop, how are PR and marketing pros monitoring AI visibility?
I posed that question to Peter Shankman’s Source of Sources (SoS). Below are some of the answers I received.
1. Manually replicating the questions prospects ask AI
William DiAntonio, founder of Brand911 takes a broad approach. He identifies common user prompts around brands and then tracks those over time.
DiAntonio said he started tracking AI visibility when his firm noticed clients would rank in search, but were “invisible when prospects used ChatGPT or Perplexity to research solutions.” His team’s efforts are “mostly” manual.
“We identify the 8-12 questions prospects actually ask before they buy (not what our clients want to rank for), then we query those across AI platforms monthly. We log who appears, in what context, and whether it’s a competitor or our client.”
Part of what the firm logs are screenshots, and they use those to track two metrics they made up called “authority displacement” and “reference content.” That means they are looking to see if their clients “are we replacing competitors in AI responses over time” visually in the screenshots and in the number of citations.
2. Tracking AI visibility around specific campaigns
Rather than try to track every possible mention in AI, Tony Crisp, who runs a marketing consultancy called CRISPx, tries to align his AI visibility tracking efforts with campaign activities.
“We’re tracking it campaign-specific, not blanket monitoring,” he said in an email. He will identify “12 high-intent queries collectors actually use during the research phase” and then “we manually query those exact phrases monthly in ChatGPT, Perplexity, and Gemini.”
“For metrics, we track mention rate (how many of our 12 queries surface the client), citation quality (are they recommended or just mentioned), and source attribution (which of our created assets—reviews, case studies, product pages—gets cited),” he wrote.
How does he package this up for a client?
“The reporting is dead simple: a spreadsheet showing query, AI platform, whether the client appeared, competitor mentions, and what content was cited.”
They also track when a buyer mentions using AI for research during sales calls. He says it happened more than 40 times for one campaign.
“The real value is knowing if you own the AI answer for your money queries before your competitor does.”
He says he spends about 90 minutes doing this analysis per client every month. The analysis has helped him make content changes that have demonstrated a “rate jump” for one client over several months.
3. Focusing on the behavior of traffic attributed to AI
One thing you can measure is AI traffic from this source. It’s easy to see in Google Analytics, for example. So instead of trying to understand mentions in AI, Scott Kasun of ForeFront Web says he’s focusing on the actions visitors take when they visit a client’s site that’s attributed to a generative AI tool.
“We started wrestling with AI visibility tracking about 18 months ago when we noticed zero-click searches eating into our clients’ traffic patterns,” he wrote. “Here’s what we’re doing differently: we’re tracking what happens after AI mentions you.”
He elaborated:
“Our tracking focuses on behavioral metrics tied to AI-influenced traffic: time-on-site, pages per session, and conversion velocity for visitors who exhibit ‘research fatigue’ patterns (they’ve clearly done homework elsewhere). We segment these users in analytics and compare their journey against traditional organic search visitors. The gap tells us whether we’re AI-ready or not.”
The thesis driving his approach is based on the idea that “getting cited by AI matters less than what your site does when visitors actually land there.”
He says one B2B client was mentioned in generative AI consistently, but at the same time, “their bounce rate spiked 40%.” They rebuilt the site’s UX and “tightened the conversion paths,” which improved their clients’ conversion rate for AI traffic by about 2.3 times.
4. Cobbling together tools and spreadsheets
Lindsey Bradshaw, a freelance PR consultant, uses a few tools and an old-fashioned spreadsheet. She runs prompts about clients across four tools – ChatGPT, Claude, Perplexity and Gemini – and documents what she finds. She lists out which companies are mentioned and how often and tracks that over time.
She strives to answer three questions…
- “Does the brand show up?”
- “How high up does it appear?”
- “How is it described (measuring tone and accuracy)?”
…across four LLMs:
“I test the same set of questions across the four LLMs, compile the answers into a table, and score them so clients can see month-over-month improvements (or declines). The scorecard usually includes scores for visibility, ranking, tone, and accuracy. I total these up into an ‘AI Visibility Index’ so there is one clear number to track over time.”
She has started experimenting with a couple of tools to augment her efforts. For example, she’s using the Answer Engine Optimization Grader, a free tool by HubSpot, “which runs its own queries across GPT-4o, Perplexity, and Gemini.”
“It measures how often each engine mentions my brand relative to competitors and produces a share-of-voice score and a downloadable report,” she added.
She also invested $99 a month in a tool called Profound, however, it only tracks ChatGPT at that subscription level. The next level up, when this blogger checked pricing, was nearly $400 per month:
“Profound’s product seems much more robust, offering a visibility dashboard and access to all of your prompts, platforms, regions, personas, and more. Still, most of the info is blocked at my price point, and setting up some of the services requires more technical savvy than I possess.”
The problem with multiple tools, she finds, “is that they provide completely different information and reporting. There aren’t many customization options for the free or low-priced tiers, and many of the cool customization features are blocked at price points that work for a freelancer like me.”
5. Determining “AI consensus”
Susye Weng-Reeder describes herself as a “PR strategist and digital creator.” The former school teacher said she’s spent 18 months studying “how AI systems index, interpret, and surface human identity.”
“I built my own manual, verifiable methodology for tracking AI citations, synthesis patterns, entity coherence, and machine-trust behavior across: Google AI, ChatGPT, Gemini, Grok, Perplexity, Claude, Bing Copilot, DuckDuckGo AI, Felo AI, and Meta AI.”
She describes a 5-step process for tracking AI visibility. She runs “repeated, structured queries (logged-in and logged-out) across all major AI models to avoid personalization.” She documents when models do certain things, like “repeating my terminology” or “citing my work.”
She then documents specific outputs from the AI:
- Signal alignment testing. She next looks to see if the various LLMs provide similar answers. “When three or more AI engines independently converge on the same biography, framework, terminology, or interpretation, I treat this as a verified signal, not hallucination.
- Changes over time. “Every test is logged with a timestamp, screenshot, and metadata note to track” outputs such as “when new citations emerge.” Recording these outputs allows her to see how AI answers change over time.
- Zero-click mentions. She tracks when “AI platforms surface my work without a user needing to click, search, or reference me directly.”
Finally, she wraps it up with a “longitudinal pattern analysis.” This is trying to determine when and where AI answers change.
It’s important to note Weng-Reeder is doing this analysis on her own work, as she has published some self-help books and pens a blog that covers many different topics, including AI.
Comment: A research proposal like this, perhaps on a shorter time frame, would have to be a separate upsell for most PR firms. And I don’t think an in-house team could invest the time it would take to document all these details.
6. Emerging tools to tell what LLMs are saying
Most of the respondents to my query were doing research by hand. Jordanne Pallese, APR, of Julep Publicity, was one of the few who responded with a tool:
“We use a platform called LLMTel to track AI visibility for our clients. LLMTel searches more than a dozen large language models (LLMs), such as ChatGPT, Claude, and Google Gemini, to see how a brand or piece of content appears in their outputs.”
The company’s website, which is very thin, says it can “scan 13+ chatbots in one click.” For Pallese, “this gives us a baseline for a metric called the Generated Prompt Results (GPR) Score, which tells us how well those AI systems recognize and mention the brand.”
There are a few other metrics her firm tracks with the tool. One includes tracking mentions in AI and “coverage analysis” which will show gaps that could stem from a publication blocking AI from crawling a site.
She says her firm does a classic before-and-after comparison to demonstrate success to clients:
“We run an initial search prior to starting work on the account to capture baseline scores and visibility profile of the brand. Alongside that, we map recent earned placements (PR hits, podcasts, articles) and correlate them with improvements in those AI visibility metrics.”
They’ll provide a report to clients that includes recommendations for improvement, which include changes to the PR strategy and suggested website optimization.
Pallese said the tool is developed by AOK Marketing (we also saw marketing and PR firms moonlighting as software companies when social media monitoring grew important). “Cost is free for the first 3 reports, then is $100 for 10 reports after that.”
Comment: The LLMTel website doesn’t provide any information on how they obtain the data. Based on the description of how the service works, it sounds to me like the tool is automating human-like prompts, so you get a snapshot in time, as opposed to continuous monitoring that you might get in a feed from a source LLM.
On the other hand, this is an actionable answer. Anyone in PR could start doing this today, where every other answer I’ve listed here would require either a heavy investment of time or significant cost in a new tool.
7. New features from SEO software tools
“We are using a tool called Semrush to track AI visibility,” said Jay Berkowitz, Founder & CEO of Ten Golden Rules, which specializes in marketing for law firms. “It has the Google AI Overview search results incorporated in the main report, and you can pull a report for ChatGPT results.”
“Despite the rapid growth of ChatGPT Google is still 10x the volume of searches, so we focus on Google AI for more opportunity,” he added.
The metrics he’s tracking are fundamental measures of marketing success – not mentions in AI:
“The metrics we track for clients are actual visits by source and leads (calls, chats and form submissions).”
Why?
“The law firm CMO wants to know how many visits came from each source, and how many leads, and most importantly, how many converted to signed clients. We need to connect to the CRM/Case Management software to connect the leads to signed clients.”
He used his instance to run a quick search on this blog, which shows how Semrush evaluates my site:
And here’s what that means for visibility AI overviews in Google:
Comment: As a traditional SEO tool, Semrush is very focused on Google’s AI efforts. I worry this is a blind spot for SEOs. Other tools outside of the Google bubble – Grok, Perplexity, Claude – are tools that test well and will take Google users over time. I think it’s likely we may see a split in search users, similar to how the news landscape fractured over time.
For its part, Semrush has a disclaimer about the complications with data veracity.
“AI search and LLM responses are fast-changing and highly personalized, which means no platform can provide exact numbers on visibility.”
At one time, personalization was new to organic search, too, so we’ve seen some of this before. Semrush is in the process of being acquired by Adobe.
Whatever you do, keep it simple
What we’ve got today to measure AI visibility is a patchwork of ideas. There are all kinds of terms being tossed around and very little consensus on meaning and definition. It’s easy to get caught up in the hype. Keep your eye out for new tools, new methods and smart presentations.
And it’s still early days. LLMs get far more headlines for being part of the AI race than they do for substance. My suggestion to PR is to keep it simple. You can spend a lot of time working out a way to measure this and not have a lot to show for it.
Chances are, particularly for media relations, the placements you earn are probably having the biggest impact on AI visibility. That’s probably a good place to start – the source of influence. At least, that’s how I’m currently thinking about this problem. That is, until the next Radian6 for AI Visibility comes along and spurs other PR software providers to tackle AI.
* * *
Subscribe by email for free:
Check out my weekly blog posts, weekly podcasts, or a monthly newsletter via Substack that rounds up interesting reading from the last 30 days (examples).
If you enjoyed this post, you might also like:
One thing AI can’t truly inform marketing or comms about is human experience
Image credit: Grok
