Overview
The agent researches your markets and fills a review inbox. It does the reading and the drafting; you do the deciding.
It works on one market at a time, using only that market's own store and language, and it never publishes anything on its own.
The two halves
The agent is two related things, and turning one on does not turn on the other.
- The proposal workflow. A configured process that researches your markets and fills a review inbox. This is what most of these pages describe.
- Agent chat. A conversational surface for asking about your own data and running work in the conversation. It is available after Agent setup and the first analysis. See agent chat.
A run
A run prepares suggestions for the markets you selected. Follow its progress on the Agent page, then review the proposed listing changes in the inbox.
Each proposal includes the context you need to evaluate it. You can keep your current wording, apply a suggestion to your draft, or edit the accepted wording before publishing.
One market at a time
Research for a market comes from that market's own store, in that market's own language. The agent does not borrow US search volume to argue for a French keyword, and it does not translate an English draft and call that localization.
What a run produces
Everything a run produces lands in the review inbox as a proposal. Spek supports App Store and Google Play listing workflows; fields depend on the selected store.
- Metadata proposals for store listing fields such as the app name and description, each with the before and after value and an explanation.
- Keyword-field proposals for the App Store, reviewed word by word, covering additions, swaps and removals.
- Screenshot proposals staged over your real screenshots. iOS only.
- Event and product page proposals carrying structured draft content. iOS only, and product pages need Apple to have made your app eligible.
None of it is live. Accepting a metadata proposal updates a Spek draft; publishing is a separate step you take. See review and apply proposals.
Common questions
How often does it run?
You start each proposal run with Generate suggestions or a direct chat request. After seven days, Spek can remind you to review fresh opportunities. Daily ranking observations do not start proposal runs.
Plan each pass around a review your team intends to complete. Daily ranking observations can provide fresh context, but they do not mean new copy was drafted or published. Check the market scope and the current run state before requesting another set of proposals.
Can it change my listing without me?
No. Everything it produces waits in the review inbox, and applying a proposal updates a Spek draft rather than the store. Publishing is a separate action you take.
Why did it propose nothing for one of my markets?
Holding position is a real outcome. If the current listing is already the best call the agent can defend, it says so on that market's card rather than inventing a change.
What does a useful review of an Agent run look like?
Work through one market at a time. Read the current and proposed field values, inspect the available keyword context and check whether the change represents your app accurately. You can keep the current text, apply a proposal or edit accepted wording in the Editor. Then review the resulting listing as a whole before deciding to publish. Completing a review does not require accepting every card, and a run that supports keeping existing metadata can still help your team avoid an unnecessary change.
What product context should my team bring to the review?
Bring knowledge of the features users can actually use, the audience you serve, the terminology your customers recognize and the brand wording you need to preserve. Store research helps evaluate search opportunities, but it does not replace a team's responsibility for truthful product claims. A phrase can attract relevant searches and still be wrong for a feature your app does not have. Check this fit in each locale, especially when a literal translation would change the promise or make the listing sound unnatural.
How should I interpret a run with few proposed changes?
First check that the run completed for the markets you requested. If it did, inspect the result and the current listing rather than judging usefulness by the number of cards. More edits are not automatically better ASO, and keeping well-supported wording can be a reasonable outcome. If research is incomplete or the run needs attention, that is a different situation and should be addressed using the displayed status. A lack of proposals alone does not establish either a problem or a guaranteed optimal listing.
Can I separate research review from the release schedule?
Yes. Research and metadata drafting can help your team prepare a change before the store version is ready to receive it. Accepted text remains a draft until an authorized publishing action sends it, and store prerequisites still apply at that point. Use the time between review and release to check localization, product accuracy and consistency across fields. If the listing changes elsewhere during that period, compare the current provider values before sending the prepared update so that the final decision uses the latest available context.