How can I check if ChatGPT recommends my brand?

Why is asking ChatGPT once not a real check?

Because AI answers change between runs, users and days, a single answer tells you very little. The founder of a firm that tests supplement categories across AI assistants told New Hope Network that asking the same question twice can return eight different brand picks, and that only 50 to 150 repetitions start to give an average that represents what the AI really recommends.

Personal context also skews results. OpenAI says ChatGPT considers your query and context such as Memory and Custom instructions when choosing products (OpenAI Help). If you run checks from a logged-in account that has talked about your brand for months, you will see your brand more than a stranger would. Use a clean account or a temporary chat with memory off.

How do you build a buyer prompt list for a DTC brand?

Build it from how customers actually talk, across five prompt types, not from your keyword list. Aim for 30 to 50 prompts; fewer than 20 gives too thin a sample, and more than 100 is hard to review by hand.

Source the wording from customer reviews, support tickets, post-purchase surveys and Reddit threads in your category. Real phrasing surfaces the constraints buyers care about (price caps, allergies, form factor) that marketing keywords miss.

How do you run the check and what should you record?

Run every prompt several times on each engine and log the same fields every time. A spreadsheet is enough.

1. Use a clean account or temporary chat on ChatGPT, and logged-out or fresh sessions on Perplexity, Gemini and Google.
2. Run each prompt at least three times per engine per month. For the prompts that matter most commercially, run more.
3. For each run, record: named or not, position in the list, the exact description of your brand, sentiment, competitors named, and every cited URL.
4. Note whether the answer showed shopping cards, a written answer or both.
5. Compute mention rate per engine (runs naming you divided by total runs) and citation share (how often your domain or pages about you are cited).
6. Keep the prompt list fixed for at least a quarter so month-to-month changes mean something.

How do you read shopping cards versus cited answers?

Treat them as two separate results with separate fixes. Shopping cards are built from structured product metadata from first-party and third-party providers; written answers draw on the model's knowledge plus web search (OpenAI Help).

• Missing from cards but named in answers: check product data. For Shopify stores, that means Shopify Catalog eligibility and completeness under Sales channels, then Agentic (Shopify Help).
• In cards but not in answers: you have a proof problem. Read the cited sources and work out why competitors appear in them and you do not.
• Wrong price or description in cards: OpenAI notes price and shipping updates can take time to reflect, and ChatGPT may generate simplified titles and descriptions from third-party data. Fix the source data, then use the in-product feedback option to report errors.
• Card labels like Budget-friendly: OpenAI says these are generated by ChatGPT from available information, not verified statements. They tell you how reviewers describe you.

Which AI visibility tool should a DTC brand use?

Start with the cheapest tool that covers the engines your buyers use, and move up only when you need citation data or more prompts. Reported 2026 entry prices:

Free checkers exist for a quick first look, such as the Semrush AI search visibility checker and Honeyb's AI visibility checker. They are fine for a snapshot, not for a baseline you will compare against.

Questions to ask any vendor: does it track product and SKU prompts or only brand prompts, does it show cited URLs, how many runs per prompt does it take, and can it be segmented to the country you sell in. And keep Google's warning in mind: no third-party tool has access to its internal ranking or AI systems (Google Search Central). For Google's own AI features, the Search Console Generative AI performance report is first-party data.

How we run this check every week, and what our own server logs showed

We run the check in this guide as a standing weekly AI Answer Lab: a fixed library of buying questions for each brand, asked in ChatGPT, Gemini and Perplexity from a US location, so the answers match what a US shopper sees. Weekly runs also add up to the repetitions a monthly read needs.

1. Lock the question library. Buying questions only (best-of, use case, comparison), worded the way customers write them. We do not edit the list mid-quarter.
2. Ask from a US location, every week. Location changes which brands and stores get named, so it is part of the setup, not an afterthought.
3. Score four things per answer: whether the brand is named, whether it is named first, which sites are cited, and share of voice against the top rivals.
4. Check cited domains for look-alikes. AI engines sometimes cite copycat stores that use a brand's name. We flag them, because a shopper who clicks one never reaches you.
5. Read share of voice as a trend. One week is noise; the line across a quarter is the signal.

What our own logs showed

Prompt checks tell you what an engine says. Server logs tell you whether it is actually opening your site. On our own site, succession.media, between September 14 and 24, 2026, ChatGPT's live browsing agent (the ChatGPT-User user agent, verified against OpenAI's published IP ranges) opened our homepage 21 times, about twice a day. One of those became a click-through: the visitor landed with a utm_source=chatgpt.com tag and booked a strategy call about a minute later.

That pairing is what we recommend to every brand: a weekly prompt baseline plus verified log and referral data. If the logs show AI agents failing to reach or read your pages, our guide on llms.txt, schema and robots.txt for Shopify covers the technical fixes. If the baseline shows rivals named and you absent, whether GEO is worth it for a DTC brand covers how to size a pilot.

What should you do when competitors show up and you do not?

Work from the cited sources, because they are the engine telling you what it trusts. For each prompt you lose, list the URLs cited and sort them into three buckets:

1. Pages you can fix: your own product, FAQ and comparison pages. Make them answer the prompt directly, in text, with evidence.
2. Pages you can earn: roundups, review sites and editorial pieces that name competitors but not you. These become your outreach list.
3. Pages you cannot touch: forums and old threads. Respond openly where appropriate and fix the underlying product or service issue.

Then connect visibility to money. Mentions matter because they send buyers: at one brand we work with, ChatGPT traffic converted at 4.99%. Pair your prompt tracking with AI-referred sessions and orders in Shopify, set up as in our tracking guide. For the fixes themselves, see how to get Shopify products recommended by ChatGPT.

If you would rather have it run for you, our AI search optimization service starts with exactly this check. In a free growth audit we run your top buying questions across ChatGPT, Perplexity, Gemini and Google AI Overviews and show you who gets named instead of you.

FAQ