Platform · Visibility by Persona
Your AI visibility, split by who is asking.
A single mention rate averages every buyer together. Visibility by Persona draws one line per audience segment you define, so you can see which buyers the engines recommend you to. Then we go to work on the segment you are losing.
Why one number is not enough
A 40% mention rate can be 70% with one buyer and 12% with another. The average looks healthy. The pipeline from that second buyer does not, and the headline number gives you no way to tell them apart.
Split by persona, the gap is a set of questions the engines answer without you, asked by a buyer you can describe. That is a content and positioning problem we can write for, fix on your site, and measure next month.
What it is
An audience segment you define, one per prompt
A persona here is a name, a colour, and a description of who the segment is: “SMB Owner, runs a ten-person team, decides on price and setup time.” Nothing is generated and there is no demographics form; you define them with us.
Each prompt belongs to at most one persona, which is what keeps the split honest: every answer we collect lands in exactly one segment, or in Unassigned, so nothing is counted twice and the segments add up to your overall rate.
- Set per prompt, in bulk, or on a whole batch promoted from Prompt Research
- Tags stay separate: a prompt can carry many tags but only one persona
- Deleting a persona keeps its prompts; their answers move to Unassigned
How it works
The same daily answers, counted per segment
Every prompt already runs daily on every engine we track, so there is nothing new to collect. The card reads the persona off each prompt and counts from there.
Assigned by hand, read at query time
The persona lives on the prompt. In the Prompts table it is a column you can sort and filter on, and a No persona filter finds the ones still waiting for one. Set persona on a selection replaces whatever was there.
Mention rate per persona
For each segment we take the answers that mention you and divide by all answers to its prompts, over the window you pick. The delta comes in percentage points, measured against the equal window before it under the same filters.
Ordered, and honest about gaps
Personas sort by mention rate, best first, and Unassigned always comes last. A persona with no answers in the window is left off rather than shown as 0%, and a delta with no previous window to compare against stays blank rather than reading as a drop.
Re-assign a prompt and its history follows
The card reads the persona off the prompt as it computes, rather than stamping it on each answer, so moving a prompt moves every answer it ever collected. We think that is the right call: the question did not change, only who asks it, and the trend should say so all the way back.
Reading the card
Winning and losing, segment by segment
Winning means a high mention rate on that segment's questions, and losing means a low one. In the example the brand appears in over half of enterprise-buyer answers and in under one in five developer answers, so the same brand in the same month has two different problems.
The card shows your own mention rate per segment, and it does not rank competitors inside a persona. For that we filter the report by persona and read Share of Voice for its prompts alone. The competitor column comes from there, and so does the difference between a gap someone else is filling and one nobody wins.
Beyond the card
One label, read everywhere the work happens
Once a prompt carries a persona, the rest of the platform reads it too.
Report filters
You can narrow any report, and any of its 24 metrics, to one or more personas alongside engines, countries, tags, and prompts. Save one report per segment and share it by link.
Content routines
Point a routine at one persona's prompts and the next article gets written for that segment. Its name and description reach the writer as the audience, which is why the description is worth getting right.
Replacement prompts
Reword or re-target a prompt and its replacement keeps the persona, while the original is archived with its history. A segment's trend carries on instead of resetting because the wording got better.
Inside our GEO work
Where it sits in how we work
This is a Discover tool: it tells us where the gap is. What we do about the segment it points at happens in Optimize.
01 · DISCOVERThis feature
The persona split turns “we are at 39%” into “we win the enterprise buyer and lose the owner-operator”. We build the rest of the month on that finding.
02 · OPTIMIZE
The weakest segment sets the next round of content and positioning work. Its prompts become the target list and its description becomes the brief.
03 · DETECT
Per-segment deltas catch a shift in one audience before it reaches the overall number or your pipeline.
Questions about personas
What clients ask about the split
Who defines the personas?
You do, with us. The platform never suggests a persona, generates one, or assigns it for you; each is a name, a colour, and a description your team and ours agree on. In Managed GEO we set them during onboarding; in GEO Consulting & Trends they follow your ICPs or buying committee. Add one, rename it, or rewrite its description from the Personas page whenever you like, and we can do the same.
Can a prompt belong to two personas?
No. A prompt has at most one persona, and setting a new one replaces the old. That is deliberate: it keeps every answer in exactly one segment, so the lines never overlap. If two audiences ask the same thing, track it as two prompts, each worded the way that audience would ask it. Tags are the other option for the overlap, since a prompt can carry many of them.
What happens when we change a prompt's persona?
Its whole history moves. The persona is read from the prompt when the card is computed, so every answer that prompt ever collected is counted under the new segment, back to the start of the window. Metrics are cached, so the card can take a moment to catch up. Deleting a persona keeps its prompts and moves their answers to Unassigned, which is why we treat a deletion as a decision rather than a tidy-up.
Can I compare competitors per persona?
Not on this card, which splits your own mention rate. To see competitors for one segment, apply that persona as a filter on the report and read Share of Voice and Position Quality for its prompts alone; the persona filter works on every metric. That is what we do in the monthly report for the segment you are losing.
How many personas should we have?
As many as you have buyers who ask different questions, which for most clients is two to five. The count is not limited, but prompts are, and each one runs daily on every engine. Give a persona enough prompts that its rate moves on real change rather than on a single answer; ten is a sensible floor. A persona with no answers in the window drops off the card until it has some.
Related features
What personas connect to
- The daily answers every persona line is counted from, prompt by prompt.
Prompt Tracking
- Point a routine at the segment you are losing; its description becomes the writer's audience.
Content Routines
- Save a report per persona, share it by link, and read every metric for that segment.
Custom Reports
- Which buying-committee roles AI recommends you to, and the pages that get each engine to repeat your positioning.
Product Marketing
See what AI is already saying about your brand.
A free audit shows you exactly where you stand across ChatGPT, Gemini, Perplexity, Copilot, Claude, Grok, and AI Overviews, and what it would take to close the gap.