Should a new product get its own property, or live under the parent brand's?

Give it its own property if the product has its own name, its own pages, and its own competitors. A shared property mixes the parent brand's existing AI visibility with the new product's, and you lose the ability to see the product's numbers on their own. A separate property also gives the launch a clean day-one baseline instead of one already muddied by years of brand history.

What if AI shows zero mentions for weeks after launch?

That's expected, not broken. AI models can only mention what they've read, and most haven't recrawled or indexed a brand-new page yet. Zero before there's indexed content and zero after don't mean the same thing. Confirm the launch pages have actually been recrawled before treating a zero as a problem worth fixing.

When should the prompts and competitors get set up, before or after the public launch date?

Before. Set up the property, the prompt set, and the competitor list while the product is still under wraps if you can. That way the very first data point AI produces after launch is already inside a tracked project, instead of the first few weeks passing before anyone starts watching.

Does this replace the standard product-marketing launch checklist?

No. It runs next to it. A launch checklist covers positioning, pricing pages, sales enablement, and press. This workflow covers one channel inside that checklist, AI search, the same way a paid media plan or an SEO plan would get its own line item.

How is annotating the launch different from just noting the date on a calendar?

An annotation lives on the same chart as the visibility data, so anyone looking at the mentions or share-of-voice trend later sees the launch date marked right where the curve starts moving. A calendar note doesn't travel with the chart, so the connection between the date and the result gets lost the first time someone looks at the data without also pulling up the calendar.

How to Launch a Product With AI Visibility From Day One

To launch a product with AI visibility tracked from day one, create a separate property for the product in Add Property before it ships, fill in a manual Brand Snapshot since there's no live page yet, generate category-entry-point and comparison prompts in Discovery Queries tagged to the launch, add the actual competing products in Competitors, and start a nightly Project with the launch date marked by an Annotation. Do this before the public launch, not after, so the first AI answers about the product land inside a tracked project instead of going unmeasured for the first few weeks.

Most product launches track everything: email opens, paid clicks, press pickup, search rankings. AI visibility usually isn't on that list, because there's nothing to measure until the product exists. That's backwards. The setup work, the property, the prompts, the competitors, can all happen before launch day, which means the day the product goes live is also the day the tracking starts, not the day someone remembers to set it up.

This guide works with any AI visibility workflow; the callouts show how each step runs in Spyglasses.

Step	What you're doing	What tells you it's working
1. Give it its own property	Separating the product from the parent brand's history	A blank property, not a shared one
2. Fill in the snapshot	Describing the product manually, pre-launch	Features, pricing, differentiators entered
3. Generate the prompts	Building the queries buyers will actually type	Prompts tagged to the launch
4. Add competitors	Listing the products it'll actually compete against	The real comparison set, not the brand's usual one
5. Start a nightly project	Turning prompts into a running, annotated project	Nightly runs live; launch date annotated
6. Watch pickup	Checking recrawl and reading early results correctly	Zeros explained by indexing, not ignored

Give the product its own property

A property is what Spyglasses tracks: one property per brand, one set of prompts, one competitor list, one visibility history. If the new product gets folded into the parent brand's existing property, its numbers get averaged in with years of the parent brand's AI visibility, and you can't pull the product's own trend back out later.

Create a new property for the product instead. Pick "Enter manually" during setup rather than pointing at a live URL, since a pre-launch product usually doesn't have one yet, or has one nobody's supposed to find. A manual property starts with a blank snapshot, which is exactly what a pre-launch product should have: nothing assumed, nothing carried over.

Fill in the snapshot before there's anything to crawl

With no live page to read, Spyglasses can't auto-fill a brand snapshot the way it would for an existing site. That's fine. Enter it manually: the features, the pricing, the differentiators, and the specific problems the product solves. This is the same information a sales deck or a launch brief already has; it just needs to live in the snapshot too, since that's what shapes the prompts generated in the next step.

Treat the snapshot as a draft that gets revised once real pages exist. The manual version doesn't need to be final copy. It needs to be accurate enough that the prompts and comparisons built from it actually reflect the product.

Pull quote: Pre-launch, AI has nothing to say about a product that doesn't exist yet. Zeros are correct then. The work is making sure they stop being correct on schedule. — “Pre-launch, AI has nothing to say about a product that doesn't exist yet. Zeros are correct then. The work is making sure they stop being correct on schedule.” — Spyglasses

Generate the prompts AI will actually ask

Once the snapshot is filled in, generate two kinds of prompts: category-entry-point prompts, the way a buyer would ask about the category before they know a specific product name, and comparison prompts, where the new product gets set against the alternatives a buyer would actually be weighing. Tag all of them with the launch so they're easy to isolate later, separate from any prompts already tracked for the parent brand or other products.

Pick prompts a real buyer would type, not ones written to flatter the product. "Best project management tool for a 10-person team" tells you something useful. "Why is [product name] the best choice" doesn't, because nobody searches that way before they've heard of the product.

Add the competing products as competitors

The parent brand's usual competitor set almost never matches the new product's. A company entering a new product category is up against different names than the ones it competes with everywhere else. List the products that will actually show up next to this one in AI comparison answers, not the brand's default rivals.

This list feeds the comparison prompts from the previous step and the share-of-voice numbers later, so getting it right matters more here than it would for a routine competitor review. A short, accurate list beats a long, generic one.

Start a project that runs nightly from day one

Turn the tagged prompts into a project and set it running nightly starting on the launch date. This is what turns a one-time snapshot into a trend: the same prompts run every night, so the visibility curve after launch is built from consistent data instead of whenever someone happens to check.

Annotate the launch date on the project the moment it goes live, and add another annotation for each coverage wave afterward, a press hit, a newsletter mention, a paid push. Those annotations are what let you look at a jump in the visibility curve three weeks later and know exactly what caused it.

Watch pickup, and read the zeros correctly

For the first stretch after launch, expect zeros. AI models can only cite pages they've actually read, and a brand-new page usually hasn't been recrawled yet, let alone indexed deeply enough to show up in an answer. A zero mention count on day two isn't a failed launch. It's a page that hasn't been fetched yet.

Check recrawl status on the launch pages specifically, not just the site generally, since a slow-moving page can sit unindexed for longer than the rest of the site. Once recrawl confirms the pages have been fetched, watch the rankings for the launch-tagged prompts and the share-of-voice numbers for the product start moving off zero. That's the signal the launch is actually landing in AI answers, not just in the press.

Signal	What a zero means before recrawl	When it's worth investigating
Mentions	Expected; nothing's been read yet	Still zero a week after recrawl confirms indexing
Rankings for launch prompts	Expected; no page to rank	Competitors rank but the product doesn't, post-recrawl
Share of voice	Expected; nothing to attribute	Flat for two+ coverage waves after annotation

Track it the same way going forward

None of this is a one-time setup. The property, the prompts, and the project stay in place after launch week the same way they'd stay in place for any other tracked brand: nightly runs, new annotations for each coverage wave, and rankings checked against the competitor set as it evolves. The only thing specific to launch is doing the setup before the product exists instead of after, so the first data point isn't also the first time anyone looked.

If the parent brand doesn't already have its own baseline, baselining its AI visibility is worth doing alongside this, since a new product's numbers read differently next to a brand that's already being tracked. Once the launch project is running, tracking AI visibility over time covers how to read the ongoing trend, and measuring AI as a marketing channel covers connecting those mentions to actual traffic. Before any of that, a free AI Visibility Report is a fast way to check where the parent brand stands today, and an AI Readiness Audit on the launch pages themselves catches indexing issues before recrawl ever gets a chance to matter.