Build AI visibility for a new product

A new product starts with no citations, and that is not a problem to be fixed on launch day. It is a baseline to be captured before launch day, so you can prove what the launch did.

What you'll learn

• What to set up before you announce anything
• What engines can actually quote about something new
• How to measure a climb from zero without fooling yourself

Step 1: Baseline before launch

Two or three weeks out, add the prompts you intend to win. They should be category and problem prompts, not product name prompts: nobody is asking for a product that does not exist yet.

Let them run. You will get a set of answers naming your competitors and not you. That is your baseline, and it is the most valuable artefact of the whole launch, because it is the only version of "before" you will ever be able to capture.

Step 2: Publish what an engine can quote

Engines quote specifics. For something new, three pages do most of the work:

1. A page that answers the category question, not a product page. "What should a [buyer] use for [job]" with an honest answer that includes you and your constraints.
2. A page that says who it is not for. The qualifier is the most-quoted sentence on most pages, and it is the one marketing removes.
3. A page with the facts in plain text. Price band, what it integrates with, where it is available. Engines cannot quote a PDF, an image, or a page that only renders after JavaScript.

Then check Readiness before you announce, because the fastest way to get nothing from a launch is to launch onto pages no crawler can read.

Step 3: Expect a lag, and know which kind

New content is not cited immediately, and the wait has two different causes you can tell apart:

• Not yet crawled. If you have the Trace worker installed, you can see AI crawler fetches. No fetch means no citation is coming yet, and the job is discoverability rather than content.
• Crawled and not used. Fetches are happening and you are still not named. That is a quality or authority problem, and rewriting the page is now the right move.

Most launch panic is the first case being treated as the second.

Step 4: Track the climb weekly, on the right number

Do not judge weeks one to four on rate. The denominator is small and the noise is larger than the signal.

Step 5: Compare against your own baseline

This is why step 1 mattered. At week eight, compare the same prompts against the pre-launch answers: who was named then, who is named now, and whether you displaced anybody or joined a longer list. Both outcomes are useful and they mean different things.

Success looks like

Your category prompts citing you on at least two surfaces, your branded prompts returning something accurate, and at least one change in the ledger at Confirmed moved against the control captured before launch.

Try this in Outercite

Before your next launch, add the prompts three weeks early and do nothing else with them. The baseline is free and it cannot be recreated later.

Related