Discovery questions versus name-recognition questions
Keep known-name questions out of discovery coverage so recognition does not inflate visibility.
On this page
Questions that include your name answer a different question from questions that leave it out. Keep both if they help you learn, but do not combine their results into one discovery number.
A discovery question asks an AI system to identify people for a problem, decision, or type of work without being told who you are. A recognition question supplies your name and asks what the system associates with it. The first tests whether your expertise appears when a prospective client might be looking. The second can reveal an outdated role, a missing public fact, or confusion with another person.
The distinction is about the question you gave the system, not a judgment about your professional quality. A name in a recognition answer is not evidence that the same person will appear in an answer to a buyer's question.
Put the two questions side by side
Consider this fictional consultant, Amina Patel. She helps UK charities plan service-design research.
| Type | Illustrative question | What it can show |
|---|---|---|
| Discovery | Which independent service-design researchers can help a UK charity plan interviews before redesigning a public service? | Whether a completed answer surfaces Amina for that work without being told her name. |
| Recognition | What does Amina Patel do in service-design research? | Whether the answer connects her name with the role and topic. |
Neither prompt is a prediction of demand, a rating, or a test of competence. The discovery question may return people other than Amina, organizations, no usable personal evidence, or an answer that needs identity review. The recognition question may repeat her name because the prompt supplied it. Report the two rows with different denominators.
Use separate reporting
Suppose Amina runs three discovery questions and two recognition questions across four engines. In a fictional snapshot, ten discovery checks return answers. Two contain a confirmed, identity-supported reference to Amina. Eight recognition checks return answers, and seven contain her name.
Her discovery observation is 2 of 10 returned discovery answers. Her recognition observation is 7 of 8 returned recognition answers. Reporting 9 of 18 as one “visibility rate” would hide the fact that the recognition prompts gave the engines the name to look for. It would make the discovery result appear stronger than it is.
If a recognition check fails or returns no answer, record it as unmeasured. A thin returned answer stays in the record but may not establish an association. Missing AI answers are unmeasured, not negative covers the denominator rule in more detail.
Gistful's question editor labels the two purposes plainly: “Find experts” and “Ask about me.” Recognition questions do not count toward discovery coverage. Keep that distinction when you export a spreadsheet or discuss a scan with a client.
Build a small question set with a job for each row
Build the set around the client situations you want to understand, then add a recognition question only when it has a clear diagnostic use.
| Question job | Example | Keep it if |
|---|---|---|
| Find an expert | Who can help a UK charity plan research interviews before a service redesign? | It describes work you genuinely provide. |
| Solve a problem | Which independent researchers can help a charity find why residents abandon an online form? | It captures a distinct buyer problem. |
| Compare fit | Which service-design researchers work with public-facing charities in the UK? | Context changes the kind of expert a buyer needs. |
| Check recognition | What work is Amina Patel known for in service-design research? | You need to inspect a specific association with your name. |
Avoid thin variations such as changing only “expert” to “consultant.” They produce more rows without adding another buyer decision. Avoid loading a discovery question with your former employer, site address, or an uncommon credential. Those details may make it a disguised recognition question.
Google's guidance for generative AI search says useful content should offer a distinct, helpful point of view and warns against producing pages for every query variation just to attract traffic. The same restraint helps here. Choose a small set of real client situations you can explain and review. Google's guide is about Google Search, not a promise about any other engine.
Read both result types without overclaiming
When a discovery answer names you, verify the identity and read the full answer. It might be a neutral mention, a list entry, or a recommendation with limits. When a recognition answer supplies a stale role, treat it as a cue to check your own public information, not proof of the system's hidden knowledge.
Save the exact text, locale, engine scope, date, and answer status for each question. Current Gistful scans cover ChatGPT, Gemini, Claude, and Perplexity. A later result is comparable only when the scope and saved discovery question match. Keep recognition evidence visible, but keep it out of the discovery baseline.