I person dozens of conversations per week pinch folks successful maturation and marketing, ranging from directors, VPs of marketing, and CMOs to SEOs successful the nitty-gritty day-to-day.
Many of my conversations impact measurement. This is an progressively challenging taxable arsenic traditional SEO metrics are breaking down pinch the advent of AI answers.
Much of the speech is spent debunking misconceptions and misguided proposal operators spot connected societal media.
It’s not an easy conversation, but it’s important. This intends telling group to debar galore of the things they spot group beforehand connected societal media. It’s not an easy conversation, but it’s the correct conversation.
My extremity present is to clear up the disorder truthful you tin necktie your AI visibility efforts to business outcomes.
What To Track
These are a operation of starring and lagging indicators that you person varying degrees of power over. I’ll make the lawsuit for why each 1 matters, past we’ll screen really to power them.
Prompts
This is the astir evident and astir important decision, because what you measurement influences behavior. From dozens of conversations, it’s besides what galore group get wrong.
It becomes the first domino successful a concatenation of mismeasurement. None of the different metrics, for illustration citation share, marque mentions, aliases visibility, matter if you’re search the incorrect group of prompts.
Most AI visibility tools for illustration Profound, Peec, and AirOps will automatically urge prompts to track, but these are seldom what you should attraction on. I haven’t confirmed this, but from what I tin tell, they analyse your website and representation existing pages backmost to prompts. They presume the pages already connected your tract are the ones that should beryllium cited aliases visible successful LLM outputs.
That might beryllium true. But astir companies we speak to say, “We aren’t appearing for the prompts we want to show up for,” which tells maine they don’t person the correct strategy and frankincense haven’t published the correct pages.
AI Visibility
You tin spot how, if you’re search the incorrect prompts, you’ll measurement visibility for the incorrect things.
From our position crossed dozens of clients, ChatGPT is the astir commonly utilized LLM, but it’s worthy search really often your marque shows up for your target prompts crossed ChatGPT, Gemini, Google AI Mode, Perplexity, Claude, and Copilot arsenic well.
This is important because ChatGPT’s personification guidelines skews consumer, while Claude’s personification guidelines skews business and enterprise. If you’re a B2B business, Claude has less users, but those users are utilizing Claude astatine work, which is the discourse that matters.
Google has besides made AI mode much prominent, and pinch Chrome having dominant marketplace share, galore group will spot AI mode outputs by default.
Tracking each exemplary separately lets you spot whether a alteration successful visibility aliases postulation is isolated to 1 exemplary aliases happening crossed each of them.
There are besides important differences successful really each exemplary searches for and presents information. For example, Claude’s usage of Brave versus ChatGPT’s usage of Bing; they person their different methodologies for selecting sources; and each exemplary has preferred sources done media partnerships.
Self-Reported Attribution
For astir of the past decade, marketers leaned connected clickstream analytics and UTM parameters to show them wherever leads came from.
That paradigm has been breaking down for a while, and it now shows an moreover smaller portion of the picture.
LLMs are zero-click by design. When personification asks ChatGPT, Gemini, aliases Claude a question, the reply is provided successful the chat. They don’t click through. They investigation successful the conversation, past they mightiness do a Google hunt for your marque straight aliases type your URL consecutive in. Your analytics and CRM platforms will log that postulation arsenic “Organic” aliases “Direct,” but the reality is that an LLM is what led to the personification going to your website.
Self-reported attribution is the simplest yet highest-signal measurement to get much of the story. We urge simply asking people, “How did you perceive astir us?” There are well-known and documented flaws to self-reported attribution (mostly quality representation and salience of touchpoints), truthful it won’t beryllium perfect, but it’s the purchaser telling you, successful their ain words, really they recovered you.
If you seizure that accusation done a section connected your lead form, you tin past way those lead sources successful your CRM to pipeline and closed revenue. That’s the eventual occurrence metric for a trading channel.
One customer that added it discovered ~5% of registrations were coming from ChatGPT contempt doing zero activity astir AI visibility.
Below is simply a floor plan from our soul dashboard. These are the number of group who filled retired our consultation shape and stated that they recovered america done an LLM, divided by the HubSpot-tagged source. Where possible, HubSpot automatically tags the root arsenic “AI Referrals” (red).
We recovered that 80-90% of leads that came successful via an AI level were incorrectly tagged arsenic “organic” aliases “direct.”
Image from author, July 2026What To Monitor, But Not Set As KPIs
These metrics thin to beryllium related but not the eventual goal, and often not what you person power over.
Citations
This is the strategical portion that group mistakenly position arsenic the occurrence metric.
Off-page sources predominate citations astatine each chimney stage. Our investigation connected citation sources recovered that, moreover for branded aliases bottom-funnel queries, 48% of sources were earned media, 30% were commercialized contented from different sites, and only 22% were from the brand’s ain website.
So you should beryllium measuring what percent of attainable mentions you presently inhabit crossed third-party surfaces. If it’s 1%, there’s monolithic headroom.
And retrieve that conscionable because you’re cited, it doesn’t mean you’re recommended aliases presented successful the LLM output.
This intends that, yes, you should look astatine citations, but not simply to spot if your website is cited, but whether your marque is mentioned connected the astir cited pages.
Sentiment
I get a batch of questions from marketers who say, “ChatGPT shows our marque successful a little favorable ray than it does our competitors. How do we amended that?”
Influencing marketplace sentiment astir a marque is simply a monolithic undertaking that nary azygous personification tin control.
There are strategies you could usage to power the sentiment LLMs present, for illustration engaging connected applicable Reddit threads aliases producing contented that paints your marque successful a amended light.
Ultimately, what LLMs coming astir your marque is not what the LLM thinks–it’s what the marketplace is saying astir your marque crossed hundreds of websites.
That has much to do pinch people’s acquisition pinch your product, your income team, your customer support, and the wide customer acquisition your institution provides. Marketing is conscionable a mini portion of that. No magnitude of trading will flooded a antagonistic acquisition pinch your brand.
So show sentiment, but dainty it little arsenic thing to power done trading and much arsenic feedback to amended the customer experience, which successful move improves sentiment.
LLM Referral Traffic
This is worthy monitoring, but not a bully KPI because you can’t power this.
OpenAI precocious changed really ChatGPT presents sources and reduced the number of sources it showed. As a result, galore websites lost ChatGPT referral traffic. However, they reversed that alteration a period later and showed more citations, starring to a ample summation successful ChatGPT referral traffic. That makes for a volatile KPI.
Beyond that, a batch of LLM postulation doesn’t get tagged arsenic LLM referral traffic by devices for illustration HubSpot aliases Google Analytics because the referral root is getting stripped. So postulation ends up being bucketed nether direct.
How To Measure The Most Important Metrics
How Do You Track The Right Prompts?
The correction galore teams make is letting their AI visibility instrumentality prime their prompts for them aliases trying to conjecture what prompts to track.
The problem is that those devices don’t cognize your customers. The bully news: you do. That knowledge already exists arsenic income telephone recordings and transcripts, onboarding calls, and customer investigation calls. These are rich | sources you tin excavation for sound of customer to understand:
- What questions travel up frequently?
- What objections travel up?
- What connection do they usage to picture symptom points aliases challenges?
If your prospects are asking these questions connected a call, they – and group for illustration them – are asking the aforesaid questions of ChatGPT aliases Claude.
The different root is self-reported attribution (see below). Once you’ve implemented self-reported attribution, your income squad tin spot whether a lead came done an LLM and simply inquire them what punctual they utilized aliases moreover petition a screenshot of the chat thread. In my experience, group are happy to share.
Between telephone transcripts and nonstop purchaser input, you switch guesswork pinch the nonstop connection customers use.
Your target prompts tin and astir apt should change. We urge reviewing prompts against voice-of-customer investigation quarterly, particularly if you’re successful a fast-moving industry.
How Do You Measure AI Visibility In ChatGPT, Claude, Gemini, Etc.?
Once you person your group of prompts, we urge utilizing Profound, Peec, aliases AirOps for search prompts.
Watch for the default channels that get tracked. For example, we don’t attraction overmuch for Perplexity, and Claude is often not included arsenic a default LLM. So we prioritize Claude, ChatGPT, Gemini, and AI Mode.
Put successful your prompts, way your competitors, and way your visibility complete time.
How Do You Set Up And Report On Self-Reported Attribution?
I mentioned supra that platforms for illustration Google Analytics and HubSpot don’t springiness the afloat image of a lead source.
We urge having a method to seizure self-reported attribution truthful that a personification tin explicitly show you really they recovered you. This won’t beryllium cleanable (attribution ne'er is), but it will springiness you different informative information point.
This will besides let you to way income opportunities, pipeline, and yet gross that came successful via LLMs.
I’ve seen this done successful 3 ways:
- Add a mobility for “How did you perceive astir us?” successful your lead aliases merchandise signup forms. You tin support it open-ended aliases connection a dropdown pinch pre-defined options and person “AI Assistant (ChatGPT, Claude, Gemini, etc.)” arsenic an option. Some clients are averse to this because they don’t want to negatively effect conversion rates. I’d urge moving an A/B trial to spot if that’s really the case.
- Add the “How did you find retired astir us?” mobility to your merchandise onboarding flow. This addresses the interest astir conversion rates, but doesn’t relationship for axenic sales-led motions.
- Have your income squad inquire connected their find calls. This requires a behaviour alteration from your income team, and the downside is it doesn’t relationship for self-service products.
Focus On What Pays The Bills
When we talk astir paid marketing, we usually don’t measurement occurrence by the number of impressions because impressions don’t salary the bills. Instead, we talk astir return connected advertisement spend. For each dollar we put in, really galore dollars do we get out?
It isn’t rather apples-to-apples, but the aforesaid logic applies. Instead of measuring visibility aliases citations (which don’t salary the bills), we should inquire whether we’re reaching the correct group done LLMs, and whether that visibility translates into business outcomes.
That intends focusing connected leads, pipeline, and revenue.
More Resources:
- How To Measure AI Search Visibility
- AI Visibility Rankings Aren’t Stable – New Research Shows It’s Mostly Statistical Noise
- We Need To Change Our Approach To AI Prompt Tracking
Featured Image: RobinRmD/Shutterstock
English (US) · 
Indonesian (ID) · 
