Look for the site's own Export or Download button first: it returns every row. If there is none, many grids load their rows from a data request you can find in the browser's Network panel and import with Excel's Power Query. Pulse Extractor's large-table export collects a grid's rows as it moves through them into a real .xlsx; in tests on a 0.5.8 pre-release build it exported 100,000 rows with every cell matching.
Why only the visible rows come back
Grids built to show thousands of rows (in admin panels, reporting tools, dashboards and many web apps) keep only the rows near your scroll position in the page, and swap rows in and out as you scroll. It keeps the page fast. It also means that select-all, copy, a screenshot, or any tool that reads the page at one moment sees a window of a few dozen rows, not the table.
Paginated tables ("page 1 of 40") and infinite scroll ("load more") are cousins of the same problem: the rest of the data is not on the page until you ask for it.
How to tell
- The grid says it has thousands of rows, but a paste into Excel gives you 20 to 60.
- The scrollbar is tiny, yet the pasted rows are the ones you could see.
- In developer tools, the number of row elements stays about the same as you scroll.
Method 1: use the site's own export
Look for Export, Download, CSV or an icon in the grid's toolbar or menu. It is the source's own copy of every row, with the right types. Some grids also let you raise the page size, for example to 500 or 1,000 rows a page, which makes Method 3 far quicker.
Method 2: get the data behind the grid
- Open developer tools (F12) and go to the Network panel. Choose the Fetch/XHR filter.
- Reload the page, or scroll the grid, and watch for a request whose response is a list of rows (usually JSON).
- If its address works on its own, in Excel use Data, Get Data, From Web with that address, then expand the list into a table in Power Query. Otherwise save the response and use Get Data, From File, From JSON.
- Check paging: a response often holds one page of rows plus a parameter for the next.
It returns clean, typed data, but takes some technical comfort, and requests that depend on your sign-in usually will not work from Power Query. Check the site's terms before reusing its data.
Method 3: page by page
For a paginated table, set the largest page size the site allows, copy or extract each page, and combine them in Excel, with Power Query's Append if there are many. It is slow but dependable for a few pages.
With Pulse Extractor
Extract Page structures the rows a page has drawn at that moment, which is the whole table on ordinary pages and the visible window on a grid that draws rows as you scroll. Pulse works on the page you are actively viewing, when you click. It does not crawl a site, follow pagination on its own or run on a schedule.
Large tables to Excel moves through a grid that loads its rows as you scroll and writes every row into a real workbook. If the page is interrupted, it continues from the rows it already holds rather than starting again.
What we tested
Synthetic test grids that render only the rows near the scroll position, opened in Chrome, exported through the same path a user takes, and checked cell by cell against the data that built the grid: every cell of every data column (a grid's row-number and selection columns hold no data). A resume means the page was deliberately minimised mid-run and Pulse continued from the rows it already held.
| Test grid | Data cells compared | Errors found | Time |
|---|---|---|---|
| 10,000 rows × 22 columns | 200,000 | 0 wrong cells, 0 missing rows, 0 duplicates | about 2 minutes (134 s) |
| 10,000 rows × 100 columns | 980,000 | 0 wrong cells, 0 missing rows, 0 duplicates | about 7 minutes (407 s), paused on purpose and resumed |
| 100,000 rows × 22 columns | 2,000,000 | 0 wrong cells, 0 missing rows, 0 duplicates | about 18 minutes (1,061 s), paused on purpose and resumed |
Version tested: a 0.5.8 pre-release build (development finished, before Chrome Web Store publication on 2026-10-05). Measured 1 October 2026.
Limits of this evidence. Test grids, not a specific third-party site; real grids differ in speed and structure. Measured on a 0.5.8 pre-release build, not re-run on the published package. An earlier unattended 10,000 × 100 run failed after the page sat hidden for 12 minutes; the rerun above passed. Long hidden stalls are a known risk.
Excel itself holds up to 1,048,576 rows per worksheet, so a 100,000-row export fits in one sheet.
Check the result
- Compare the row count with the grid's own total.
- Check the first and last rows, and a few in the middle.
- Look for duplicates where pages or scroll windows meet.
- Make sure numbers and dates arrived as numbers and dates, so they sort and add up.
When Pulse is not the right tool
- The site has an export or an API. Use it: it is the source's own copy of every row.
- You need the table refreshed on a schedule. A scheduled tool or the site's data feed suits that; see Browse AI vs Pulse Extractor. Pulse works when you click.
- A grid that draws rows as you scroll, today. The current release captures the rows on screen; use Methods 1 to 3 until the large-table export is published.
Limitations to know
- Time. Large grids take minutes, not seconds, with any in-browser method.
- Authorization. Being able to see a grid does not by itself give you the right to export it; see Responsible Use.
- Protected browser pages. Chrome does not let extensions run on its own protected pages, such as browser settings or the Chrome Web Store, so Pulse is unavailable there. That is a browser restriction, not a Pulse setting.
Questions
Why does Ctrl+A only copy some of the rows?
Because the grid only keeps the rows near your scroll position in the page. The others are drawn when you scroll to them, so they are not there to copy.
How many rows can Excel hold?
A worksheet holds up to 1,048,576 rows and 16,384 columns.
How long does a large export take with Pulse?
In tests of the large-table export on a 0.5.8 pre-release build, about 2 minutes for 10,000 rows by 22 columns, about 7 minutes for 10,000 rows by 100 columns, and about 18 minutes for 100,000 rows by 22 columns. Real grids vary.