How to Baseline Your Brand's AI Visibility

Jim Wrubel

Jim Wrubel

8/4/2026

#Agencies#AI Visibility Baseline#How-to#Workflows#AI Search Visibility
How to Baseline Your Brand's AI Visibility

To baseline a brand's AI visibility, run one scan across every major AI platform (ChatGPT, Gemini, Claude, Perplexity, and Google AI Overviews), then correct the auto-built brand profile before anything else reads from it. That corrected, verified snapshot is the baseline: a fixed point every future report gets measured against.

A new client signs, or a prospect takes your call, and the first real question is the same either way. How visible is this brand in AI search right now? Not a guess from typing one prompt into ChatGPT last week. An actual number, checked across the platforms buyers use, that you can put a stake in the ground with.

Skip the baseline and every future report compares against nothing. This guide walks through the five steps that turn a domain name into a complete baseline: add the property, run the scan, fix the profile the report builds itself around, read the results the right way, and hand over something the client or prospect can actually use. It works with any AI visibility tool; the callouts show how each step runs in Spyglasses.

Key takeaways

  • A baseline is a fixed point, not proof of progress. It's a corrected, verified snapshot every future report gets measured against.
  • Test every platform in one run. ChatGPT, Gemini, Claude, Perplexity, and Google AI Overviews disagree with each other, so a manual check of one tool misses real gaps.
  • Correct the brand profile first. It feeds every prompt and mention match generated after it, so a wrong category or stale differentiator skews everything downstream.
  • Read for the gap, not the score. A named competitor beating you on specific questions is a stronger pitch than an isolated percentage.
  • Package it under your own name. A raw dashboard export doesn't close a deal; a branded PDF or deck does.
StepWhat you're doingWhat tells you it's working
1. Add the propertyCreating a property for the brand, or a batch of prospectsProperty appears with a scan already queued
2. Run the scanTesting the brand across ChatGPT, Gemini, Claude, Perplexity, AI OverviewsReport status moves to complete
3. Correct the profileFixing the auto-built brand summary before anything downstream reads itCategory, audience, and differentiators all match reality
4. Read the baselinePulling share of voice, consistency, and the competitive gap out of the noiseYou can name the single biggest gap in one sentence
5. Package itExporting a branded, client-ready versionA PDF or deck is ready to send under your name

Add the property you're baselining

Everything starts with a property, the record a baseline gets attached to. For an existing client this is a five-second step. For new business it's usually more than one brand at a time, since a pitch rarely rests on a single competitor comparison.

If you're evaluating a list, don't add domains one at a time. Paste the list, run it as a batch, and let every prospect scan on the same schedule so the numbers you're about to compare were all pulled under the same conditions. That consistency matters more than it sounds; a report run today and one run three weeks ago aren't measuring the same AI landscape.

Run a full AI visibility scan

One scan tests the brand across every major AI platform at once instead of you opening five tabs and typing the same question five times. That matters because platforms don't agree with each other. A brand can be well described on Gemini and nearly invisible on Perplexity, and a manual spot check across one or two tools will never catch that gap.

A single property's scan usually finishes in a few minutes. A batch of prospects takes longer, sometimes a few hours, since each one runs its own full set of checks rather than sharing a queue. Kick it off, move on to something else, and come back when it's done instead of watching a progress bar.

Correct the brand profile everything else builds on

Here's the step agencies skip, and it's the one that costs the most later. When a scan runs, it crawls the site and writes a working profile of the brand: category, audience, differentiators, competitors. Everything generated after this point (the prompts tested, the mentions matched, the competitors compared against) reads from that profile.

The profile is a good first draft, not a finished one. A crawler can miss a recent repositioning, misread a multi-product company as single-category, or lift a differentiator from stale homepage copy. None of that is a dealbreaker, but it's worth five minutes to check before you build a client narrative on top of it. Fix the category if it's wrong. Sharpen a differentiator if the site undersells it. Confirm the audience is who the brand actually serves, not who they're hoping to serve someday.

Pull quote: A baseline nobody corrected is just a photo of whatever the crawler guessed. Fix the profile before you trust the number.
A baseline nobody corrected is just a photo of whatever the crawler guessed. Fix the profile before you trust the number.Spyglasses

Read the baseline like an analyst, not a scoreboard

A finished report has a lot on the page: consistency score, share of voice, per-platform breakdown, competitor rankings, citation sources, a list of recommendations. Reading all of it top to bottom for every prospect isn't a baseline, it's an afternoon lost. Go in with a question instead: where is this brand losing ground, and to whom?

Start with the competitive comparison. A brand can look fine in isolation and still be trailing badly the moment you check who AI mentions ahead of it for the same questions. That gap, named specifically, is worth more in a pitch than any score on its own. "You're at 17% share of voice" is a number. "A named competitor gets mentioned first in 3 out of 5 comparison questions" is a reason to hire you.

Package the baseline for a client or prospect

A raw dashboard export doesn't close a deal or hold up in a client kickoff. Whoever receives the baseline needs something that reads like a deliverable: your agency's name on it, not a vendor's, and a format they can open without a login.

Two exports cover most situations. PDF for anything going over email or into a shared drive. Slides for anything landing in front of a room, a pitch meeting or a client's quarterly review. Set the branding once and it applies to every export after, so this isn't a step you repeat per client.

The baseline is the starting line, not the finish

A baseline doesn't prove anything moved. That's not its job. Its job is giving you a fixed point, corrected and verified, that every future report can be measured against. Add the property, run the scan, fix the profile before it feeds anything downstream, read for the gap that actually matters, and hand it over looking like your own work.

Once a client is signed on the strength of a baseline, the next move is turning that single snapshot into something that runs every night. Tracking AI visibility over time picks up exactly where this guide ends, with the competitor list, prompt coverage, and monitoring project a real engagement needs. If the brand's AI visibility has never been checked at all, a free AI Visibility Report is the fastest way to get the first number on the board.

How to Baseline Your Brand's AI Visibility