What is an AI brand mention worth?

Three things people call an AI brand mention

The word “mention” hides three different events. They have different evidence, different durability and very different value.

• A citation. The answer links to your page. The cited URL is the evidence, and a citation sample can record it repeatably.
• A recommendation. The answer names your brand without linking. The only evidence is the response text, which means you need a sample of prompts you control.
• A comparison. The answer places you next to alternatives. Same evidence problem as a recommendation, with more context to interpret.

At scale, only the citation is verifiable through a search API. That is why the measurable layer starts with citations, referrals and crawler access instead of a total mention count.

What is the price of an AI brand mention?

A mention is not an ad slot. There is no inventory, no impression count and no stable unit to price. An answer is generated for one question to one user; the engine publishes neither how many people saw it nor how often it will be generated again.

The query also changes the value. A mention in “what is [your brand]?” mostly reaches people who already know you. A mention in “best [category] for [use case]” reaches someone choosing. A vendor quoting one price per mention is averaging those two into a number that means nothing for either.

If a vendor does quote a per-mention price, ask what prompt set and run count it buys. You are pricing a sample inside that vendor's method, not a placement in ChatGPT or Google. The quote can be compared with another sample quote; it cannot be compared with an ad CPM or an organic visit.

What is the value of an AI brand mention?

The value shows up downstream, and it can be estimated as a floor:

AI referral sessions × conversion rate × value per conversion

Use the first-party referral record, not GA4 alone: referrers get stripped, so the console count is a floor rather than a total. Then compare those sessions with the rest of the site. If AI referrals convert at the same rate as organic search, the mention is worth the same as an equivalent organic visit — no premium required.

Branded search demand is a second-order signal. A lift in branded queries after an answer appears is worth watching, but it cannot be attributed to one mention; treat it as context, not a conversion.

How to tell a real mention from a proxy

• Evidence: a citation sample with the cited URL, or a referral record with its landing page.
• Proxy: branded search lift, a crawler fetch, or one person asking ChatGPT and screenshotting the answer.

Proxies are useful for forming a hypothesis. They are not counts, and a report that mixes them with citations will overstate what it knows. Keep the evidence and the proxy in separate columns — the AI traffic report does exactly that.

Why sentiment scores are unreliable

There is no published per-brand sentiment API. A model generates an answer from the sources it retrieves and the patterns in its weights, and the same prompt can produce a different tone in a different session. A vendor sentiment score is a sample of prompts run through a classifier; the classifier is the product, not the engine.

Use sentiment as a directional input when you can see the sampled answers. Do not put it in a dashboard next to clicks as if the two numbers shared a unit.

A practical way to value a mention

1. Pick 10–30 buying-intent queries where an answer could plausibly name you.
2. Sample AI Overview citations for them; keep the cited URLs, not just a yes/no count.
3. Record AI referral sessions per landing page with a readiness check confirming the pages are fetchable.
4. Compute referral conversion and value; label it a floor, not a forecast.
5. Track branded search demand separately and never attribute it to one answer.
6. Re-check crawler access after every deploy, so a lost mention has a timestamp.