Personas

AI does not give every buyer the same answer. Ask for the best accounting software as the CFO of a 15-person startup and you may get a different shortlist, and different warnings, than the controller of a 2,000-person company gets. Personas let you see those differences. A persona is one synthetic buyer. Add personas to a project and every discovery prompt in it also runs once a day as each of them, so you can compare how AI answers that buyer with how it answers the same prompt asked with no persona.

What You'll Learn

• What a persona is, and exactly what we send to AI
• How to generate personas from your customer segments or create them by hand
• How to add personas to a project, and how their runs count against your plan
• How to read Perspectives: baseline comparisons, thin samples and who AI recommends to each buyer
• How to filter the Recommendations page by persona

What a Persona Is

A persona describes one buyer in three parts:

Personas are shown role first, for example "CFO, early-stage startup". A name is optional. When you add one, it appears next to the role for your team only and is never sent to AI.

What we send

We turn the fields into one short first-person paragraph of at most 600 characters and put it in front of the prompt, word for word. For a CFO persona and the prompt "What is the best AP automation software for a growing company?", the platform receives:

I'm the CFO at a B2B SaaS company in the United States with about 120 employees. I sign off on the purchase. We use QuickBooks Online and Gusto. My top priorities: close the books in under five days. Constraints to keep in mind: no dedicated AP clerk. What is the best AP automation software for a growing company?

The question itself never changes. Only the asker does, which makes the persona answer a fair comparison with the same prompt asked with no persona.

The editor shows this text live under What we send, with a character count. If you would rather write it yourself, turn on Write what we send myself: we send your text (up to 600 characters) instead, and the fields still group results in Perspectives. The name, role group and segment are never part of what we send; they only change how a persona is shown and grouped.

Creating Personas

Open Personas in the property sidebar, next to Prompts. Creating, generating and editing personas is free, and you can keep as many as you like. Nothing runs until you add a persona to a project.

Generate them from your customer segments

Personas are drafted from your Brand Snapshot, so a sharp snapshot makes sharper personas.

1. Click Generate personas on the Personas page, or Create personas on the Customer segments tile of the Brand Snapshot.
2. Pick a customer segment.
3. Choose what to draft:
   • One persona: one buyer who represents the segment, a quick first read.
   • Stakeholder panel: 3 to 5 people at one company across the buying committee, such as who signs, who runs the review and who uses the product every day.
4. Review the drafts. Keep the ones that fit, edit anything, then create them.

Nothing is saved until you create the drafts. Drafts never list your own brand or your tracked competitors among the tools a buyer already runs, so the persona does not steer answers toward or away from them.

Create one by hand

Click New persona and fill in the fields. The What we send preview updates as you type.

Edit a persona

You can change any field at any time. Only a change to the text we send starts a new version. When the persona runs on a project, saving starts the next version from the next daily run, and Spyglasses adds an automatic annotation to each of its projects so trend charts mark the day the wording changed. Earlier answers stay under the earlier version. Changing the name, role group or segment does not start a new version.

Archive and restore

Archiving a persona stops its daily runs after today. When the persona is on a project, you are asked to confirm first, and the dialog lists the projects it runs on. Its answers so far stay in Perspectives. To bring it back, open Show archived and restore it; its daily runs resume on the same projects.

Personas belong to the property. A location uses its parent's personas.

Adding Personas to a Project

Choose personas in the Ask as personas section when you create a project, or later on the project's Personas tab. The picker groups your personas by role, so you can select a whole role group in one click.

Once personas are on a project:

• Every discovery prompt in the project also runs once a day as each persona. Brand-identity and comparison prompts are left out, because a prompt that names you invites an answer about you.
• Persona runs use the AI platforms your property tracks, except Google AI Overviews, so ChatGPT and Gemini by default, plus any add-on platforms your property has. AI Overviews answers a search, not a buyer, so a persona version would repeat the regular run.
• The first answers arrive today. Adding a persona starts its first run right away. After that it runs with your daily tracking.
• Persona runs use the brand-wide run only. They are not repeated for each location.

Persona answers feed Perspectives and the persona filter on Recommendations. Goals and citations come from the answers asked with no persona.

How Persona Runs Are Counted

Each prompt asked as one persona is one tracked prompt from your plan's allowance, the same way a location is. A project with 12 discovery prompts and 2 personas adds 24 runs a day. When a persona already asks a prompt through another project, those runs are not counted twice, and the picker names the other project.

Before you save, the Daily runs panel in the picker shows the change in one line, for example:

+24 runs/day: 12 included, 12 billed, about 12 credits per 30 days

A meter under it shows how many tracked prompts you will use after saving against the number your plan includes.

• Within your plan: the runs are included and you can save straight away.
• Over your plan: the panel says how far over you will go and what it costs. Each run a day beyond your plan uses 1 credit every 30 days (1 credit per 30 daily runs, prepaid). Tick the confirmation checkbox to save.
• Unlimited plans: every run is included.

Removing a persona from a project stops its runs there after today. Its answers so far stay in Perspectives.

Your Headline Metrics Stay the Same

Dashboard numbers count only answers asked with no persona. We call these answers the baseline. Share of voice, mentions, consistency, citations and the Recommendations page (with no persona selected) all read the baseline, so adding or removing personas never moves them. The coverage and data accuracy figures on a project's Prompts tab cover the baseline runs too.

Perspectives

Perspectives shows how AI answers your prompts when a specific buyer asks, compared with the baseline. It is a dashboard page next to Recommendations. Open it with Open Perspectives on a project's Personas tab, from a persona's card on the Personas page, or with Compare every persona with the baseline on the Recommendations page when a persona filter is on.

Same question, different asker

Each row is one persona, or a group of personas, compared with the baseline over the same prompts and platforms in the same window. If a persona runs on 3 of your 50 prompts, its baseline is those same 3 prompts asked with no persona. Google AI Overviews is left out on both sides.

Filters and controls

The categories behind these rates are explained on Recommendations.

What the page shows

• The biggest difference. The page leads with the largest clear difference for your brand in one sentence, with the buyer's rate, the baseline and the difference. It adds the main reason AI gives when it cautions, or the phrase AI most often uses when it calls you best for this buyer, and links to the answers behind it. When no buyer differs clearly from the baseline, the page says so rather than promoting noise.
• How each buyer sees your brand. One row per buyer. The dot is the buyer's rate, the dark tick is the baseline over the same prompts, and the whisker is the buyer's 90% interval. Rows are sorted by difference, and color appears only where the difference is clear. Pick a row to see who AI recommends to that buyer.
• Who AI recommends to this buyer. Share of recommendation among the selected buyer's answers: each brand's recommendations (top choice or one of many) as a share of every brand's recommendations, untracked brands included, next to the same brand's baseline share. Your brand is marked You and brands you don't track are marked Not tracked. Choose Baseline (no persona) to see the market with no persona.
• Compare every brand. Closed by default. The chosen rate for your brand, your tracked competitors and the most-named untracked brands, as each buyer. Differences show as an arrow and points, with no green or red, because a competitor's gain is not good news for you.

Reading a baseline comparison

A difference is the buyer's rate minus the baseline rate over the same prompts, in percentage points. If AI names you its top choice in 12% of the answers an early-stage CFO gets and in 31% of the baseline answers for the same prompts, the difference is 19 points lower. Points are not percent: a move from 20% to 30% is 10 points.

Every difference comes with a 90% interval. A difference is clear only when its whole interval sits on one side of zero. Anything else could be noise, and the page draws it in gray.

Thin samples

A buyer with fewer than 30 answers in the window is an early read. The row is grayed out, shows how many answers it has (n), and is never marked clear. At that size one answer moves a rate by several points. Wait for a few more days of daily runs, or widen the window to 90 days, before you read it.

As a guide, a persona on 5 prompts across two platforms gets about 10 answers a day, so it passes 30 answers in about three days.

Brands in the buyer's stack

Answers often restate the tools the asker says they already use. When a persona runs QuickBooks Online, an answer may name QuickBooks Online without recommending it. The AI judge rates a tool the buyer already uses as a generic mention unless the answer recommends it for the question, and Perspectives marks those brands as in this buyer's stack, so a restated tool does not read as a recommendation.

Before the first answers

• First answers today: a persona added to a project today has its first run under way. The page fills in as the answers come back, next to the baseline for the same prompts.
• No personas yet: the page explains what Perspectives compares and links to the Personas page, where you can generate personas and then add them to a project.

In the client portal, Perspectives is read-only: clients can read every comparison and open the answers behind it, and the links to manage personas are not shown.

Filter Recommendations by Persona

The Recommendations page and its answers page have a Persona filter. Choose a persona to narrow the rates, shares, cautions and answers to the answers asked as that buyer. Links from Perspectives land here already filtered to that persona or group.

While a persona is selected, a line above the page says whose answers you are reading and links back to Perspectives. Google AI Overviews is not offered as a platform, because persona runs skip it. With the filter on None (baseline), the page shows the baseline, exactly as it does without personas.

Opening a persona answer shows the persona and a What we sent section with the full text the platform received: the persona's paragraph followed by the unchanged prompt.

Best Practices

Start with a stakeholder panel for your main segment

A panel shows how AI treats each seat on the buying committee for the same purchase. Differences between the economic buyer and the technical evaluator are often the first useful finding.

Keep each persona specific and true

Real titles, a real company size and the tools that segment actually runs produce the answers that segment actually gets. Vague personas get vague answers that look like the baseline.

Give personas enough prompts

A persona on a handful of prompts takes longer to pass 30 answers. Add personas to a project whose discovery prompts cover the questions that buyer asks.

Read the cautions by persona

When Perspectives shows a buyer is cautioned against you more often than the baseline, filter Recommendations by that persona and open Why AI cautions against you. The reason and audience AI gives are the claims to answer for that buyer.

Edit wording sparingly

Each change to what we send starts a new version and marks the charts. Settle the wording early, then leave it alone so the trend stays comparable.

FAQ

Do personas change my dashboard numbers? No. Headline metrics count only answers asked with no persona, so adding, editing or removing personas never moves them.

Does creating personas cost anything? No. Creating, generating and editing personas is free. Only persona runs on a project count as tracked prompts.

Why don't persona runs include Google AI Overviews? AI Overviews answers a search query, not a buyer, so a persona version would repeat the regular run. Perspectives leaves it out of both sides of the comparison.

Is the persona's name sent to AI? No. Only the paragraph under What we send goes in front of the prompt. The name, role group and segment stay in Spyglasses.

Why is a buyer grayed out in Perspectives? It has fewer than 30 answers in the window. Wait a few more days, or switch the window to 90 days.

What happens when I archive a persona? Its daily runs stop after today, and its answers so far stay in Perspectives. Restore it any time and its runs resume on the same projects.

Explore personas with an AI assistant (MCP)

If you've connected the Spyglasses MCP server, list_personas returns a property's personas with the exact text we send for each, and get_persona_lens returns the same baseline comparison as Perspectives. See MCP → Account data.

Related

• Recommendations: The categories behind every Perspectives rate, and the persona filter
• Projects: Add personas to a project so its prompts run as each buyer
• Brand Snapshot: The customer segments personas are generated from
• Prompts: Author the discovery prompts personas ask
• Annotations: The chart markers written when personas are added, edited or archived
• Locations: How location runs are counted, the same way as persona runs