Choose fields according to the dataset you need: use From a list for repeated records, Just text for individual values, and consider Table Studio when the data is already visible and requires no interaction. Before building the robot, decide what each output column should mean, which records and pages are in scope, and whether the robot must run again later.
Start with the shape of the data
Browse AI’s current guidance distinguishes repeated records from individual page values. A search-results page with many similar products is a different extraction problem from a single page with a title, price, and contact details. Pick the capture mode that matches the structure you want returned, not simply the number of visible elements.
As an Amazon Associate I earn from qualifying purchases.
| What the page looks like | Best starting point | Why |
|---|---|---|
| Search results, a directory, product grid, reviews, or another repeating pattern | From a list | Captures similar items as rows with consistent data points and supports pagination. |
| A page with scattered, one-off values such as a title, price, contact details, or specifications | Just text | Lets you select individual elements and label them as fields or columns. |
| Data is already visible and needs no clicks, typing, or login | Table Studio | Browse AI says it can propose a structured table that you can review, add columns to, or edit before saving. |
| Values appear only after a click, dropdown, form entry, or login | Robot Studio interaction, then the appropriate capture mode | Train the interaction that reveals the values before capturing them. Complex multi-page journeys may need workflows or multiple robots. |
Browse AI’s “From a list” guidance also says a robot can combine list extraction, text capture, and screenshot capture when one task needs more than one kind of output. For example, a list may be the main dataset while a separate value or visual record is also needed. See the help pages on From a list vs. Just text and building your first robot.
Decide what belongs in the output before training
Write a short schema before opening the builder. Each field should represent a value you actually need, and its name should make sense to someone reading the resulting table later. Browse AI’s best-practices guidance recommends planning the fields, their structure, the page or row scope, update needs, and how you will use the extracted data.
#1 Best Overall
- Purpose: What question will this dataset answer or what task will it support?
- Fields: Which values are necessary? Avoid selecting every visible item by default.
- Meaning: Use labels that distinguish similar values, such as “monthly price” and “annual price” when both appear.
- Scope: Which records, pages, or page types must be included?
- Inputs: Include a run parameter, such as a search term, only if different runs need different inputs.
- Refresh: Will the robot run once, or repeatedly to track changes?
There is no established universal ideal number of fields. The useful test is whether each column serves the intended output.
Build and review the robot
- Open the page containing the data. Choose the actual page where the values appear, not necessarily the site’s homepage. Browse AI’s getting-started guide uses a pricing page rather than a homepage as the appropriate target for collecting competitor pricing.
- Choose the capture approach. Use Table Studio as a starting point for data that is visible without interaction; use From a list for repeated records, or Just text for individual values.
- Select the intended records or elements. For list extraction, wait until the dotted outline encloses exactly the repeating records, then select it. Review the proposed structure. If it misses a required field or groups the page incorrectly, use manual selection and label the values.
- Name columns by meaning. Make the label specific enough to tell similar values apart. In Table Studio, Browse AI’s guide describes naming a column and describing the content it should contain; it also describes deleting unwanted columns.
- Set the list size and pagination behavior. Match the setting to how the site exposes additional records; the correct choice is part of defining the dataset, not a minor setup detail.
- Preview representative results. Check several rows, confirm each value appears under the intended label, and decide whether any blank cells reflect genuine differences in the source.
- Save only after checking scope and shape. Confirm the table has the records, columns, and context needed for the way you plan to use it.
For the documented workflow, see extracting data From a list, building your first robot, and Browse AI’s best practices for extraction and monitoring.
Set pagination to match the page
For From a list extraction, choose the behavior that actually advances or reveals the next records. Browse AI documents these settings:
| Page behavior | Setting |
|---|---|
| Next button, arrows, or numbered pages | Click next |
| A “Load more” or “Show more” button | Click load more |
| Additional items appear as the page scrolls | Scroll down |
| All required records are already visible | No more items |
Some sites use JavaScript navigation that looks like ordinary pagination but behaves like a load-more control. If a run stops before reaching the intended records, inspect what the page actually does when you advance and select the matching option. The pagination guide covers From a list extraction; it does not cover other capture modes or traversal of individual detail pages. See Browse AI’s pagination instructions.
Interpret missing values and table structure
An empty cell is not automatically a robot failure. Some records on a site may lack a value that other records have—for example, a rating may be present for only some products. Inspect multiple records to tell source variation from a selection error.
Browse AI’s list guidance says automatic structure can omit a necessary field, and that the documented choice is full automatic or full manual selection. If a required field is missing from the automatic result, switch to manual selection and select all the fields you need rather than assuming you can add just that one to an automatic selection.
Browse AI describes individual captured text values as columns and list records as rows in a list tab. It also describes context columns such as extraction date and input parameters. These can matter when interpreting results across repeated runs: a value may be meaningful only alongside the date or input that produced it. See Understanding your data structure and the capture-mode comparison.
Quick wins for a faster PC:
Repair Windows errors before they cause bigger problemsFix Now →Scan for outdated or missing drivers - takes under a minuteDriver Scan →When the source is a table or hides details
Browse AI says it can detect many HTML and visually styled tables. If a row expands to reveal additional information, first decide whether the visible values are enough. If not, train the interaction that reveals the hidden data before capturing it. For complicated nested structures, the table guide notes that manual selection may be needed.
Best Value
Browse AI’s table-extraction guidance is available at How to extract data from tables on a web page.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Or skip the browser setup
If the task is to capture a webpage screenshot rather than build a structured Browse AI dataset, ScreenshotNeo is a screenshot API and MCP server from Yorker Media. One GET request returns a PNG, JPEG, WebP, or PDF. For example, a cURL request can save a WebP screenshot:
curl -G "https://api.screenshotneo.com/v1/shot" -d access_key=YOUR_API_KEY --data-urlencode url=https://stripe.com -o shot.webpSee the ScreenshotNeo API documentation for request options. Its clean-shot flow accepts cookie or consent banners like a visitor and removes more than 60 known consent platforms, newsletter popups, and chat widgets; each step can be turned off. Bot checks or CAPTCHAs, blank pages, timeouts, failed loads, and cache hits are not billed, with response headers indicating the page verdict and billing status. An MCP server provides take_screenshot, get_page_info, and capture_pdf tools for Claude, Cursor, and other MCP clients. The Free plan includes 1,000 screenshots per month with no card; paid plans start at $5 for 3,000.
Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minutePC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Sign up free for 1,000 screenshots a month—no card required.
Troubleshooting field selection
- The list outline includes too much or too little: Recheck that you selected the repeated records themselves, not a surrounding page region. Review the proposed dataset, then switch to manual selection if the automatic structure is unsuitable.
- A required column is missing: The automatic list structure may not recognize it. Browse AI’s documented workaround is to use full manual selection and select all required fields.
- The run stops before all records are collected: Confirm whether the site uses a next control, a load-more button, scrolling, or no additional records. Change the pagination setting to match the observed behavior.
- Some cells are blank: Compare several source records. If the site itself omits the value for some records, blank cells can reflect real variation; if the source shows the value, review the selection and field mapping.
- A value appears only after interaction: Train the click, dropdown choice, form entry, or login needed to reveal it before choosing the capture mode.
- A table is nested or reveals row details: Decide whether the visible cells meet the need; otherwise reveal the details during the robot flow. Complex nesting may require manual selection.
These behaviors and limitations are described in Browse AI’s guides to list extraction, pagination, and webpage tables.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
