How to test whether ChatGPT recommends your product — and fix it when it doesn’t.

The fastest way to start testing whether ChatGPT recommends your product costs nothing. Open a fresh, incognito-mode ChatGPT conversation — prior context and memory can bias answers — and ask the category, problem, comparison, and local-intent questions your actual customers use. If you sell project management software, don’t start with “tell me about [your brand].” Ask instead: “What’s the best project management tool for remote engineering teams under 50 people?” Record the exact prompt, the date, and every brand named in the answer.

The critical discipline is repeating the same prompt set across multiple sessions and days. Because ChatGPT’s recommendations are non-deterministic — two users asking the same question can get different answers based on retrieval variability — a single snapshot is not a reliable signal. Run at least five sessions per prompt and look for which brands appear consistently. That pattern, not any individual answer, is your real baseline. Document which sources ChatGPT cites when it does name a competitor: those pages are your content gap list.

Profound is purpose-built for what marketers now call GEO — Generative Engine Optimization. It runs your defined prompt set automatically every day across ChatGPT, Perplexity, and Gemini, records every mention of your brand, scores the sentiment, and surfaces which third-party sources the models are pulling from when they name you or a competitor.

The real value is trend data. A single prompt result is a snapshot; 30 days of daily results shows whether your recent content or structured-data improvements are actually moving your AI visibility score. Profound also surfaces the competitor brands consistently appearing in the same answer space, which gives you a concrete list of the editorials, Reddit threads, and review pages you need to appear in. For teams serious about GEO, it removes the manual logging overhead that makes the free approach unsustainable at scale. You still need to act on what it finds, however — or pair it with a tool that fixes the gaps automatically.

Scrunch AI goes a layer deeper than simple mention tracking. It analyses how ChatGPT and other large language models frame your brand when they do name it — is it described as a market leader, a budget option, a risky choice? That framing shapes buying decisions just as much as presence does, and most monitoring tools miss it entirely.

In our testing, Scrunch surfaced cases where a brand was technically mentioned in ChatGPT answers but consistently qualified with negative language pulled from a single outdated review thread. The fix in that case was not more content — it was addressing the source of negative sentiment at the origin. That level of diagnostic depth is worth the higher price tag for brands where AI-generated perception is a commercial priority. For brands just starting to measure AI visibility, the free manual approach or Profound is a better first step.

ChatGPT’s shopping experience runs on Bing’s merchant index, not Google’s. That means the product feed you submit to Microsoft Advertising is the primary source ChatGPT uses when a user asks it to find, compare, or buy a specific product. If your product feed is absent, outdated, or missing required attributes, you are structurally invisible in shopping conversations regardless of how good your website content is.

Setting up the feed requires a Microsoft Advertising account, website verification, and a product data file with required attributes: id, title, description, link, image URL, availability, price, brand, GTIN or MPN, and condition. Brands already running Google Shopping campaigns have most of this data — it is largely a matter of submitting the same feed in Microsoft’s format. Treat feed hygiene as a continuous process: stale pricing or out-of-stock items that remain live can actively suppress recommendations. For a deeper look at how AI shopping signals overlap with paid strategy, see our guide on connecting AI to your Google and Meta ad feeds.

Structured data is the machine-readable layer that tells AI models exactly what your page is about, what the product costs, what reviewers said, and who the brand is — without the model having to infer it from prose. ChatGPT’s retrieval layer cross-references entity signals across multiple sources; a well-structured product page is one of the clearest signals that your brand is a credible, verifiable entity worth recommending.

Run Google’s Rich Results Test and Schema Markup Validator on your top product and category pages. The most valuable schema types for ChatGPT visibility are Product (with offers, aggregateRating, and review properties), FAQPage, and Organization with sameAs links to your profiles on Trustpilot, G2, and major directories. Every field left blank is context the model cannot use. Read our guide on generative engine optimization fundamentals for a full schema implementation checklist.