To check whether AI recommends your B2B company, start with the questions a buyer would ask before knowing your name. Repeat them, retain the answers and compare your business with named competitors. A single result is a clue, not a dependable score.
Ask a buying question, not a branded question
“Tell me about our company” tests how an engine describes a known business. “Which providers help a B2B services firm respond to enquiries faster?” tests discovery. Both can be useful, but they answer different questions.
Use your actual category, buyer and constraints. Start with questions your customers ask in sales conversations; avoid loading the prompt with your brand name when testing unprompted discovery.
Keep mentions, recommendations and citations separate
- Mention
- Your business is named in the answer.
- Recommendation
- The answer presents your business as a suitable option.
- Citation
- The answer links to a source. A source may mention you without being your website.
A citation to your article is not necessarily a recommendation to buy your service. A recommendation can also appear without a link to your site.
Repeat the test and record the conditions
Keep the wording, provider, date, search mode and account context with each answer. Use fresh conversations and repeat each question. Compare like with like rather than treating a changed prompt or provider as evidence that visibility improved.
Repeated answers can vary, as SparkToro’s research illustrates. A small manual check is useful for spotting a possible gap; it cannot establish your visibility across all buyers.
Look at who appears instead, and why
Record the competitors named and the sources cited. Check whether your own public material clearly states your category, intended buyer, use cases and scope. Review relevant directory profiles and independent coverage for omissions or outdated descriptions.
These are areas to investigate, not proof of what caused an AI to select a competitor. Avoid assuming that matching one competitor’s page will produce the same result.
Check access before rewriting everything
A blocked or unindexed page may not be available to search systems. Review robots rules, indexing settings and hosting restrictions. Search access and model-training access are distinct: OpenAI documents separate controls for OAI-SearchBot and GPTBot.
Connect visibility to the rest of the buying journey
Being named is only the first step. A buyer still needs to understand your offer, find credible information and receive a useful response. Check positioning and enquiry handling alongside shortlist presence so improvements address the whole route to a deal.
When is a fuller assessment useful?
When you have a named category, identifiable competitors and enough deal flow for consideration to matter, a structured assessment can separate a visibility issue from positioning or conversion gaps. The GTM Readout combines repeated shortlist measurement with the wider go-to-market diagnosis and written corrections.
See the complete Readout methodology, or use The Shortlist Questions as a starting point.
Sources and limits
- SparkToro’s original research on recommendation variability.
- OpenAI’s crawler documentation.
- Google’s generative AI search guidance.
This guide is a practical starting point. It does not report client results or guarantee inclusion in any AI answer.