Most founders check whether ChatGPT recommends their brand by opening a tab, typing a prompt, and scanning the response. If they see their name, they relax. If they don’t, they panic.
Both reactions are wrong. A single ChatGPT check tells you almost nothing.
Why a single ChatGPT check gives you a false answer
When someone asks ChatGPT for a recommendation, it gives a direct, curated answer typically mentioning 3 to 5 brands. If you’re not one of them, you’re invisible. There’s no page two.
But here’s the problem with checking once: there is a less than 1% chance of an AI generating the same recommendation list twice. The brands that appear, the order they appear in, and the framing around them all shift between responses. A single manual check captures one snapshot of a system that never holds still.
This makes the whole exercise misleading. You might check and see your brand, conclude everything is fine, and miss that you only appear in 8% of relevant responses. Or you might check, not see your brand, and assume you’re invisible when you actually appear in 40% of queries phrased differently.
The scale compounds the stakes. ChatGPT users send about 2.5 billion prompts per day, a growth rate that took Google a decade to achieve. Those AI answers are shaping perception long before anyone lands on your site.
How ChatGPT actually decides which brands to mention
ChatGPT doesn’t search the internet the way Google does. It pulls from a combination of sources: its core language model (trained on massive amounts of public internet content up to a cutoff date), real-time browsing in premium versions, structured data like Google Business Profiles, online reviews, directories, and authoritative content it encountered during training.
If you’re already investing in AI search visibility, the good news is that traditional search performance feeds AI citations directly. According to Ahrefs data cited by SEO Works, 71.7% of ChatGPT’s citations come from pages with organic search presence. And 76.1% of citations in Google AI Overviews come from pages already ranked in the top 10 organic results.
Seer Interactive’s analysis of over 300,000 keywords found a 0.65 correlation between ranking on page one of Google and being mentioned in ChatGPT’s responses. Interestingly, backlink quantity showed a weak relationship, suggesting that brand strength plays a bigger role than link volume. A separate Ahrefs study of 75,000 brands found that branded mentions, branded anchor text, and branded search volume all correlated significantly with appearances in AI-generated content.
Brand signals matter more than link-building volume. If people search for your name, mention it in forums, and reference it in articles, AI models are more likely to surface you. This is one of the key AI ranking factors to understand.
Step 1: Build your prompt list before you check anything
Before you open ChatGPT, figure out what your buyers are actually typing into it.
This is different from keyword research. People talk to ChatGPT in full sentences: “What’s the best project management tool for a 10-person remote team?” rather than “project management software.” The prompts that matter are the ones real buyers use during their decision process.
Ask current customers how they found you via AI. Add a field to your contact form encouraging customers to share the prompts they used. Talk to your sales team about the questions prospects ask before they buy. These real-world inputs are more valuable than guessing.
Build a list of 10 to 20 prompts across three categories:
- Discovery prompts: “What’s the best [your category]?”
- Comparison prompts: “How does [your brand] compare to [competitor]?”
- Problem prompts: “[Specific problem your product solves]”
This prompt list becomes the foundation of every check you run.
Step 2: Run the manual check, and understand its limits
With your prompt list ready, open ChatGPT and start querying. For each prompt, note:
- Whether your brand appears at all
- Where it appears in the response (first mentioned, middle of a list, afterthought)
- How it’s framed (recommended, mentioned neutrally, compared unfavourably)
- Whether ChatGPT cites a source for mentioning you
Do this for every prompt on your list. Then do it again with slightly different phrasing. A minimum viable manual process means running 3 to 5 variants of each prompt, because with less than a 1% chance of repetition, any single run is unreliable.
This is tedious. It’s also the free version of checking. It will give you a rough directional sense, but it won’t give you the data you need to make decisions or track progress. For that, you need the metrics.
Step 3: Measure the three metrics that matter
Traditional rank tracking is irrelevant for AI answers. There’s no position 1 through 10. The right framework, as outlined by Profound, centres on three metrics:
-
Visibility score: the percentage of AI responses that mention your brand for a given set of prompts. This is your core number. If you appear in 15 out of 100 responses, your visibility score is 15%.
-
Share of voice: how your visibility compares to competitors for the same prompts. If a competitor appears in 60% of responses and you appear in 15%, your share of voice gap is clear regardless of any single check.
-
Average position: where your brand typically appears within answers. Being the first brand mentioned carries different weight than appearing fourth in a list of five.
Together, these tell you how visible your brand is and reflect its relative performance in your industry. They also give you something a single manual check never can: a baseline to measure improvement against.
For a broader view of how to approach this, see our guide on AI visibility tracking.
The tool landscape: what to use and when
Manual checks don’t scale. Several purpose-built tools now exist to monitor ChatGPT brand mentions, and the right one depends on your budget and needs.
| Tool | Best for | Starting price |
|---|---|---|
| PromptRush | Prompt-level tracking with alerts | $19/mo per project |
| Otterly.AI | Budget-friendly ChatGPT mention tracking | $29/mo |
| Profound | Enterprise share of voice and citations | $99/mo |
| Semrush | Existing Semrush users adding AI visibility | $99/mo |
When evaluating tools, prioritise ones that provide full AI answer text (not just mention flags), exact placement of brand mentions inside responses, and historical visibility data to track change over time.
One thing these tools share: they solve the randomness problem by running prompts repeatedly and aggregating results, which is exactly what manual checking cannot do at scale.
Why not just use your existing SEO tools? Because ChatGPT’s sources have only a 39% overlap with Google’s sources. Traditional SEO rank tracking tools cannot tell you if your brand is visible in AI answers. AI search has its own ranking system, and tracking it requires tools built for that system. This is the core difference between traditional SEO and AI search optimisation.
What being invisible actually costs you
ChatGPT only recommends 1.2% of local businesses, while Google’s local results show 35.9% of locations. AI visibility is up to 30 times harder than Google visibility.
The audience is already there: 39% of US consumers have already used generative AI for shopping-related tasks, and more than half (53%) plan to use it to research products, get gift ideas, compare pricing, and discover brands.
The danger is that drops in LLM brand visibility often stay invisible until traffic, demand, or referrals decline. By the time you notice in your analytics, you’ve already lost months of potential discovery. Google still holds 93.39% of global search market share versus ChatGPT’s 0.44%, but customer journeys are becoming multi-modal. A buyer might start with ChatGPT, validate on Google, and convert on your site. If you’re missing from step one, you may never reach step three.
The signals that lift your chances of being recommended
If your checks reveal low visibility, the fix isn’t to game a prompt. It’s to build the brand signals that AI models weight heavily.
Volume of mentions matters. It takes approximately 250 unique mentions or publications for an LLM to form a definitive understanding of your brand. Not 250 backlinks. 250 distinct references: articles, forum posts, reviews, directories, podcast transcripts, anything that creates a unique textual mention.
Ratings are a filter, not a bonus. AI models act as gatekeepers. If your average rating falls below roughly 4.0 stars, your brand may be filtered out of AI results entirely. This isn’t a ranking factor you can optimise around. It’s a threshold you need to clear.
Organic search performance feeds the cycle. With 71.7% of ChatGPT citations coming from pages with organic search presence and a 0.65 correlation between page-one Google rankings and ChatGPT mentions, investing in answer engine optimisation pays dividends across both surfaces.
Brand search volume is the leading signal. Branded mentions, branded anchor text, and branded search volume all correlated significantly with AI appearances in the Ahrefs study. People searching for your brand by name is one of the strongest indicators that AI models should mention you.
Checking if ChatGPT recommends your brand is not a one-time exercise. It’s a measurement practice. Build the prompt list, establish your baseline across the three core metrics, and track changes monthly. A single check is a coin flip. A systematic process is a strategy.