> ## Documentation Index > Fetch the complete documentation index at: https://docs.rippit.com/llms.txt > Use this file to discover all available pages before exploring further. # How research tables work > Research a topic or a list of items on the web, across AI assistants, or for people, and collect every finding in a table Rippit keeps up to date. A research table collects outside information into Rippit. You describe what you want to learn, pick how Rippit should search, and every finding lands as a row you can query, chart, enrich with an [AI table](https://docs.rippit.com/enrich/ai-tables.md), or point an Agent App at. To create one, open **Tables**, choose **Add table**, then **Create Research Table**. Or describe the research in any chat, and Rippit sets it up in that conversation. ## Research data integrations Pick a tool to see how it works. ### Web and People Data > **[Exa](https://docs.rippit.com/research/exa.md)** > **[Firecrawl](https://docs.rippit.com/research/firecrawl.md)** > **[Clay](https://docs.rippit.com/research/clay.md)** ### Model Provider AEO Data > **[ChatGPT](https://docs.rippit.com/research/chatgpt.md)** > **[Claude](https://docs.rippit.com/research/claude.md)** > **[Gemini](https://docs.rippit.com/research/gemini.md)** ## The three search methods Each research table can use one method or several. All of them share the same brief, inputs, schedule, and results table. | Method | What it does | What each row holds | Tools | | - | - | - | - | | Web search | Finds web pages that match your brief | Title, link, date, source, a summary, and the page text | [Exa](https://docs.rippit.com/research/exa.md) by default | | People search | Finds people who match your brief, and can look up their work email and phone number | Name, role, company, location, profile link, why they match, and any contact details you asked for | [Firecrawl](https://docs.rippit.com/research/firecrawl.md) by default, [Clay](https://docs.rippit.com/research/clay.md), or [Exa](https://docs.rippit.com/research/exa.md) | | Agent search | Asks AI assistants a set of questions with their own web search turned on, and records how they answer | The question, the answer, citations, the searches the model ran, and a full trace | [ChatGPT](https://docs.rippit.com/research/chatgpt.md), [Claude](https://docs.rippit.com/research/claude.md), [Gemini](https://docs.rippit.com/research/gemini.md) | Use web search to track pages and news, people search to build a list of prospects or contacts, and agent search to see how AI assistants talk about a topic, a brand, or a competitor. ## Write the research brief The brief is the instruction every search follows. Rippit drafts one from your request, and you can edit it at any time. A good brief says: * **What counts as a relevant finding.** "Product launches announced by the company," not just "news." * **Which sources you trust**, and which to skip. * **What evidence a result needs**, such as a publish date or a direct quote. * **What to leave out**, and what to do when a result is uncertain. Don't put a time range like "last 7 days" in the brief. Set how far back to look in the table's settings instead (see [Schedule](#schedule-and-history)). ## Research a topic or a list A research table works in one of two ways: * **One topic.** The brief alone drives the search. "Find recent reviews of AI support tools." * **One search per item.** Pick a column from an existing Rippit table, or upload a CSV, and Rippit runs the brief once for each value in it. Add `{{value}}` to the brief where the item should go: "Find product launches from `{{value}}` in the last quarter." For a per-item table you can: * Research every value in the column, pick specific values, filter the rows first, or keep only the top values by another column (for example, the 50 largest accounts by revenue). * Add **context columns**, so each search also sees other details from that item's rows. * Research up to 100 values per table. CSV files need a header row and must be under 5 MiB. The column you split on is fixed once research starts. Search methods and the brief can still change. ## Sample, then run 1. **Save** the brief and settings. 2. **Run a sample.** A sample is a small, separate preview on up to three of your inputs. It doesn't touch your live results or schedule. Check the findings on the **Sample** tab. 3. **Run now** when the sample looks right. A full run uses every input and writes to the live results table. A sample is capped on findings and research time, so it shows the kind of results you'll get, not full coverage. Changing the brief, the methods, or their settings marks older samples as out of date. ## Schedule and history A research table can run **daily**, **weekly**, or only when you click **Run now**. Turning on a schedule starts at the next scheduled time, not right away. For web search, set how many days back the first run should look. After that, each run looks back only to that input's last successful run, so you collect new pages without duplicates. Leave it empty for evergreen research with no date limit. ## Use the results Results build up in one table. Web pages and people accumulate across runs. Agent search keeps every study, and the results show the most recent one. From there you can: * Ask Rippit questions about the results in chat. * Build an [AI table](https://docs.rippit.com/enrich/ai-tables.md) on top of it to classify, score, or summarize each row as it arrives. * Use it as the source for an [Agent App](https://docs.rippit.com/plg/agent-apps.md) that reports on it every week. The output columns of a research table are fixed. To add your own columns, use an AI table. ## Cost Research uses AI credits for the model doing the research, plus provider costs for the search tools. People search contact lookups use extra provider credits. The table shows the reported cost for each run and the total across live runs. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.