What it will not do

The boundaries below are deliberate. Each one exists because the alternative would produce a confident answer that is wrong, and a wrong answer costs more than a refusal.

It does not change anything you own

The Reach Agent reads. It does not edit your listings, rewrite your titles, adjust your prices, or touch your store. When it recommends a title change, it gives you the title and you decide.

It does not spend money

Campaign work is draft mode only. It produces ad groups, keyword lists with search volume and cost per click, bid ranges, and budget splits. It never pushes to an ad platform, never commits spend, and never modifies a live campaign. Everything it writes is a draft for your review.

It does not tell you what to charge

Pricing is research. It shows you the real products other independent stores sell that match yours, what those stores charge, and how an individual product's price has moved over time.

It will not tell you that you are priced too high or too low, and it will not give you an average, median, or spread for a product segment. The reason is specific: segments span ranges too wide for a middle to mean anything. All-cotton bed sheets run from $20 to $1,500 inside a single segment, and a central figure across that range describes no real product. You get the actual products with their actual prices, and you draw the conclusion.

Price history for an individual product is unaffected. That is a real series about one item.

It does not answer catalog questions without a catalog

"Which of my products are invisible to search" has no answer until the agent knows what your products are. Rather than answering about a market instead and letting it read as though it were about you, the agent says the question needs your catalog.

Each account's catalog is isolated. One person's products never appear in another person's answer.

It does not fill gaps

If the data does not support an answer, the answer says so. It will not estimate around a hole, round a missing figure into a stated one, or invent a date for an undated count. An empty result is reported as an empty result.

It does not show you its plumbing

Answers do not carry raw query results, table names, internal schema, or stack traces. You get findings in plain language with the sources named. When something fails on our side, you get a clear statement that it failed rather than a technical dump.