Web research

Web research uses OpenAI to search public sources. A brief can combine several competitor, market, and custom research sections, each with its own questions and scope. When a question names an app on the App Store, Zentrik also reads its recent reviews directly; if several apps match, it asks which one. No provider credentials are needed.

1. Open Discovery → Research, then New research. Choose Competition, Market, or Public feedback, enter the subjects, and choose Review research plan. Each starts with editable questions. For a specific investigation, choose Custom research, enter or dictate a question, and choose Continue.
2. Review the questions, source scope, and research window. If several App Store apps match, choose the app you mean. The plan shows the confirmed apps and sample reviews. Choose Refine to change the full plan or remove an app you do not want to read.
3. Choose Start research. The results stay in the brief for review before you add them to Discovery.
4. Open a finding to read its answer and supporting sources. Select a citation to inspect the saved observation. Open the Sources view, then a source group to search all saved items, filter reviews by app or stars, or show the items cited by findings.
5. Choose a destination product and Add to Discovery. The addition uses the sources saved in that run.
6. After the addition completes, choose Only when I ask, Weekly, or Daily. Automatic research starts only after you choose and save a cadence. Use Updates beside the report to change frequency, pause, or resume an active brief. Schedules managed by Zentrik remain read-only.

Choose a baseline to understand current capabilities or needs, including relevant older and undated documents. Choose recent developments for a dated window of 1–90 days. Subsequent scheduled runs look for new sources, with a three-day overlap. Repeated URLs are skipped; this does not detect edits to existing pages.

Workspace owners and admins can create and manage saved briefs in Settings. People with workspace edit access can run research from a permitted brief. The questions and scope are sent to OpenAI. Research frequency is separate from Brief emails.

New sources become Signals for the selected product. A web page keeps its link, dates, and an AI summary. A review keeps its author’s words, rating, date, and link. Company announcements represent the company’s claims.

Find saved research by topic, question, or product from the research home. Open a report. The Findings view shows each answer in a short finding; select it for the full answer and the reviews and pages it cites. Use What changed or Needs evidence to focus the review. When subjects have matching questions, Compare places their saved answers side by side; choose a question and open an answer for its evidence. From the research library, choose Compare research to bring together any number of saved subjects with shared questions. Each subject shows the date and scope of its saved answer; choosing subjects does not start new research. If an answer cannot be prepared, the saved sources remain available. Sources groups what the run read, such as App Store reviews or web pages; open a group to search and filter every saved item. When a run reads at least five new reviews of an app, What reviewers raise groups them into problems, requests, and what reviewers value, with how many reviews raise each; select one to read exactly those reviews. A later run of the same plan starts with what changed: which answers changed, how many reviews it read, and which themes more or fewer reviews raise. A changed plan starts a new comparison. Earlier runs are listed by date; choose one to read it. Owners and admins can manage the active brief’s frequency under Updates, beside the report.

Repeated source URLs and reviews already added to Discovery are skipped. An unadded preview does not move the saved review position. Research can return no results when no new sources meet the brief. Search results do not guarantee complete coverage of a market.

Sources per section limits qualifying source documents. A first preview uses one search round per section and defaults to three sources. Scheduled runs can make up to three rounds to fill the target. Rejected sources and duplicate URLs do not fill it. All searches retain the questions, date window, and source restrictions. A source target is not a guarantee that matching evidence exists.

A run can read up to four confirmed App Store apps, sharing a budget of 40 reviews with at most 25 per app. These are recent reviews from the US storefront; the feed does not establish where authors live or which customer segment they belong to. An explicitly restricted App Store section uses its confirmed reviews without a separate web search. A completed run lists Coverage notes when some topics remain unverified. They distinguish gaps in coverage from a researched topic with no new findings.

Individual customer reviews can qualify even when their hosting page has no publication date. Review dates stay attached to their observations. A month-only date is kept as a month and qualifies when the entire month is inside the research window. If a section returns no accepted sources because candidates were rejected or coverage was limited, the next run keeps the previous research window. The window is also retained when a section falls short of its source target and coverage remains limited.

Use Refine this research below the findings to describe a correction, then choose Review revised plan. Review the change before choosing Run revised research. To edit the full brief, choose Review and run again. Review and add the new sources before applying the revised brief to an existing schedule. Pause an active brief to stop automatic runs.

Pause a brief before archiving it. Archiving preserves its run history and previously captured Signals.

You can also review and manage research through a connected MCP agent or the Research API. Read access, permission to run research, and permission to manage updates are separate. Review the plan and destination before starting or adding research; automatic updates require your explicit choice.