What counts as an AI visibility baseline?

A completed AI Visibility Report with a reviewed Brand Snapshot. The report alone gives you scores; the reviewed snapshot confirms those scores are measuring the right brand, category, and competitors. Skip the review and you have a number, not a baseline you can trust.

How long does the first report take to run?

Four to fifteen minutes, depending on how many pages Spyglasses needs to crawl and how many discovery queries it runs across platforms. Bulk imports of 20 prospects at once can take a few hours to fully clear, so kick off a batch and come back to it rather than waiting on it live.

Do I need a paid plan to run a baseline report?

No. A free preview covers roughly half the data from just a domain name, which is enough for an early prospect conversation. The full report, at $15 standalone or included in a subscription, adds the per-prompt breakdown, full competitor list, and citation sources you need for a real client baseline.

What's the difference between a baseline report and ongoing tracking?

A baseline is one report, a single snapshot of where a brand stands today. Tracking runs the same kind of prompts nightly inside a project and watches for real movement over weeks. You need the baseline first; it's the number every later report gets compared against.

How often should I re-run the baseline itself?

You shouldn't need to, once tracking is running. The baseline report is a starting line, not a recurring task. If a property sits idle for a few months, a fresh report (free quarterly on active plans) gives you a clean new starting point before you restart tracking.

How to Baseline Your Brand's AI Visibility

To baseline a brand's AI visibility, add the property in Spyglasses, run an AI Visibility Report, and correct the Brand Snapshot it generates before you read a single score. In four to fifteen minutes, the report crawls the site, queries ChatGPT, Gemini, Claude, and Perplexity twice each (once with web browsing, once without), and scores brand consistency on a 0-100 scale, share of voice, and competitive position across five starter discovery queries. Every one of those scores is measured against the snapshot, so get the snapshot wrong and the whole report is measuring the wrong brand.

This is the step agencies skip when they're moving fast, and it's the step that makes the rest of the work defensible. A tracking project, a monthly client report, a new-business pitch; all of it eventually points back to this first number. A brand consistency score of 74 out of 100 only means something if the category, audience, and competitors behind it are the ones the client actually recognizes. Get any of that wrong at the crawl stage and nothing built on top of it holds up.

This guide walks through establishing that baseline for a new client or prospect, correcting the parts of the process that need a human eye, and getting the report in front of the people who need to see it. The callouts show how each step runs in Spyglasses, but the sequence works with any AI visibility tool that starts from a crawl.

Step	What you're doing	What tells you it's working
1. Add the property	Giving Spyglasses a domain to crawl and score	Property (or Prospect) shows up in the dashboard
2. Run the first report	Crawling the site and querying AI platforms	Report status changes to Completed
3. Review the Brand Snapshot	Correcting category, audience, and differentiators	Every snapshot field matches how the client describes itself
4. Read the report	Working through the scores in the right order	You can explain the consistency score and share of voice in plain language
5. Share the report	Getting the baseline in front of the client or prospect	A link, PDF, or PowerPoint is sent, not just viewed internally

Add the property to Spyglasses

Every report needs a property behind it: a domain Spyglasses knows to crawl. For an existing client this is a straightforward add, a root domain plus an optional location. For a prospect you haven't signed yet, treat it as a Prospect instead of a full client property. Prospects live in a separate view on the organization dashboard, so a pile of cold outreach domains doesn't clutter the client list you actually manage day to day.

If you're pitching more than one account at a time, bulk-add the list instead of adding domains one by one. Paste a list of domains (with an optional location after a comma), and Spyglasses validates every row before anything runs, flagging invalid domains or ones that already exist as properties. A single import caps at 20 prospects, and a full batch can take a few hours to clear since each one kicks off its own report.

Run the first AI Visibility Report

Once the property exists, run the report. Spyglasses crawls the homepage plus several other high-value pages to build a Brand Snapshot, then queries ChatGPT, Gemini, Claude, and Perplexity twice each, once with web browsing allowed and once without, so you can see what the AI already "knows" about the brand versus what it can find live. A run takes four to fifteen minutes depending on site size and how many discovery queries it generates.

The first report for a new property is free on an active Brand, Agency, or Enterprise plan, and any property gets one more free refresh every three months after that. Run it again sooner and it costs credits, so this is one report worth getting right the first time rather than re-running out of impatience.

Review and correct the Brand Snapshot

The crawl builds a Brand Snapshot: company name, category, target audience, key features, problems solved, differentiators, and pricing model, pulled straight from the site. It's usually close. It's rarely perfect, and every score in the report downstream depends on it being right.

A restaurant chain's crawl might read "Korean Restaurant" when the client thinks of themselves as "Korean BBQ, fast-casual." A B2B tool might get tagged with the wrong ICP because the homepage leads with a feature list instead of a customer description. None of that is a bug. The crawler is reading what's on the page, and pages don't always say what a founder or brand lead would say out loud. That gap is exactly why this step exists.

Go through every field against how the client actually describes their own business, not just what the site says. Fix the category, confirm the audience, and fill in differentiators the crawl missed entirely; that field is often blank because it takes a human read to know what actually sets a brand apart from its category, not just a scrape of the homepage copy.

“A baseline isn't a report you file away. It's the only number every future report gets compared against, so it has to be measuring the right brand.” — Spyglasses

Read the report the right way

A completed report has a lot in it: a consistency score, share of voice, per-platform performance, a competitor breakdown, discovery query results, grounding search rankings, and citation sources. Read in the wrong order, it's a wall of numbers. Read in the right order, it answers one question after another.

Start with the consistency score. It measures how accurately AI describes the brand compared to the corrected snapshot, and it sets the frame for everything after it: a brand AI barely recognizes needs a different plan than one AI already understands well but doesn't recommend often.

Consistency score	What it means	What it usually calls for
70-100	Strong accuracy; only minor gaps in features or differentiators	Focus on share of voice and citations, not correction
50-69	Core description is right, but key details (audience, features) are off	Fix site content AI is misreading, then re-check next report
0-50	Brand is missing from AI training data or confused with another brand	Site accessibility and content depth are the first things to check

From there, move to share of voice (how often the brand shows up when someone asks a relevant question), then the per-platform breakdown, since a brand can score well on ChatGPT and barely register on Perplexity. Competitors and discovery queries show who's winning the same searches and how. Recommendations come last on purpose; they only make sense once you understand what's actually wrong.

A baseline that stays in your dashboard doesn't do anything for a client relationship or a pitch. Get it in front of the person who needs it, in a format they can act on without logging into Spyglasses themselves.

Every report comes with a shareable web link, plus PDF and PowerPoint exports built for email or a deck. Agency-plan accounts can white-label all three, so a client-facing export carries your firm's branding instead of a generic dashboard screenshot. For a prospect, this is often the entire pitch: a specific, branded read on their current AI visibility that no capabilities deck can match.

Now you have a number to build on

A baseline isn't the finish line. It's the starting line, and the whole point of running it carefully is that everything you do next gets measured against it. A tracking project shows movement instead of noise. A monthly report shows progress instead of a fresh, unrelated snapshot. A pitch shows a specific gap instead of a generic promise.

Get the property added, run the report, fix the snapshot before you trust the scores, read them in the order that actually explains them, then put the result in front of the person who needs to see it. Once that baseline exists, the natural next move is turning it into nightly tracking so you can prove whether the work you do afterward actually moves the numbers. And if you're running this baseline on a prospect rather than a signed client, turning that first report into the pitch itself is the next step worth reading.