---
title: "Google Removed num=100: What It Means for Rank Tracking"
description: "Google stopped honouring the num=100 parameter in September 2025. Rank trackers now load ten pages per keyword and Search Console impressions fell."
url: https://proxynet.io/blog/google-num-100-parameter
date: 2026-11-19
author: "Acar Diveroli"
category: "Use Cases"
lang: en
---

# Google Removed num=100: What It Means for Rank Tracking

It is a Monday in mid-September 2025 and you open the Search Console report for a client's site. The impressions line has fallen off a ledge on desktop, yet the average position line has jumped upwards, as if the site had climbed in the rankings over the weekend. Clicks have hardly moved, and the agency's rank-tracking tool shows gaps beyond page two. Nothing on the site changed. Google had stopped honouring a small piece of its search URL: `&num=100`.

This post explains what the parameter did, when it stopped working and what Google said about it, why Search Console impressions and average position moved, what it means for the cost and depth of rank tracking, how to read your own data across that date, and what is worth tracking now.

> **Note: Short answer**
>
> `num=100` was a URL parameter that made Google show up to 100 results on one page instead of ten. It stopped working around 11-12 September 2025, and Google told Search Engine Land on 18 September 2025 that the parameter "is not something that we formally support". Rank trackers that read the top 100 in one request now need about ten page loads per keyword. In Search Console, many sites saw desktop impressions fall and average position improve at the same time: the measurement changed, not the visitors. Read your data across that date with clicks first and the date noted, and take your ranking data from Search Console or its API, not from automated Google queries.

## What was the num=100 parameter?

The `num` parameter told Google how many results to put on one results page, and `num=100` asked for a hundred. A normal search URL looks like `google.com/search?q=running+shoes`, and the results arrive ten at a time. Adding `&num=100` to that address returned positions 1 to 100 as one long page.

For a person, it saved nine clicks on "Next", and Google once offered the same choice in its search settings. According to [Search Engine Land](https://searchengineland.com/google-search-confirms-it-does-not-support-the-results-per-page-parameter-462244), Google removed that setting from the interface in 2018, but the URL parameter kept working for years afterwards.

For rank-tracking software it mattered far more. A rank tracker is a tool that records where a website appears in the results for a list of keywords, usually every day. With `num=100`, one request returned the whole top 100, which made deep tracking cheap.

## When did num=100 stop working, and what did Google say?

It stopped working around 11-12 September 2025, without an announcement from Google. The timeline comes from search-industry press, because Google did not publish a notice of its own:

1. **12 September 2025.** Search Engine Roundtable reported that Google appeared to be testing the removal: the parameter worked for its editor about half the time. An update to the same article said it had then stopped working completely.
2. **15 September 2025.** Search Engine Land wrote that rank and position tracking was "a mess", that third-party tools now paid about ten times as much to collect the same data, and that Search Console reports looked different too.
3. **18 September 2025.** Search Engine Land published Google's reply. A Google spokesperson said: "The use of this URL parameter is not something that we formally support." The same article dates the switch-off to late on the Thursday of the previous week, which was 11 September 2025.

Google did not explain why it made the change. Press articles speculated about scrapers and AI products that read Google's results, but those are guesses, not Google's words. The safe reading is simple: the parameter was never a promise, and it no longer works.

## Why did Search Console impressions drop after the change?

Many sites lost a large share of desktop impressions overnight, and the most widely discussed explanation is that part of them had come from automated tools, not people. Google has not confirmed it, so treat it as a theory that fits Google's own counting rules.

Search Console's help page on [impressions, position and clicks](https://support.google.com/webmasters/answer/7042828?hl=en) says an impression is counted whenever a link to your site appears on the current page of results, whether or not the searcher scrolls down to see it. Position is the highest place your link held on that page. Apply those rules to a hundred-result page:

1. A tool loads a results page with `num=100` for a keyword.
2. Your site sits at position 64 on that long page.
3. Because the link was on the page, an impression could be counted at position 64, even though no human scrolled that far.
4. Thousands of such loads add impressions at deep positions, which inflates the total and makes the average position number larger, that is, worse.
5. Once `num=100` stopped working, reaching position 64 took seven pages of ten, tools went less deep, and those impressions disappeared.
6. Impressions fell and average position improved, because the remaining impressions sit higher on the page.

Search Engine Land's [15 September report](https://searchengineland.com/google-search-rank-and-position-tracking-is-a-mess-right-now-461984) described exactly this pattern: SEO consultant Brodie Clark and others saw a significant decline in desktop impressions together with a sharp improvement in average position. Clicks should not move with this kind of change, because a tool that loads a page does not click on it.

There is a second factor to keep in mind. Google's Search Console [data anomalies page](https://support.google.com/webmasters/answer/6211453?hl=en) lists a separate logging error that kept Search Console from accurately reporting impressions from 13 May 2025 until 27 April 2026. Google says only impressions, CTR and average position were affected, not clicks, and that the fix may show as a decrease in impressions. So any comparison that crosses 2025 and early 2026 has two measurement changes in it, not one.

## What changed for rank tracking and Search Console?

The table compares the situation before and after September 2025.

| Item | Before | After |
|---|---|---|
| Results on one page | Up to 100 with `num=100` | 10 |
| Page loads for the top 100 | 1 | About 10 |
| Desktop impressions | Included many deep positions | Fewer |
| Average position | Worse, pulled by deep positions | Better, closer to page one |
| Clicks | Counted only when a person clicks | Same |
| Tracking depth in tools | Top 100 in one request | Depends on the tool |

The last row is not a fixed rule. Depending on the vendor, the answer can be fewer positions by default, less frequent checks of deep positions or a higher price for the top 100, so check what your own tool does now.

## Why does it matter for the cost of rank tracking?

Deep tracking became about ten times more expensive, because every extra page of ten results is one more request. A tool that tracks 10,000 keywords to position 100 used to make 10,000 page loads a day; ten results at a time, it needs roughly 100,000. More requests mean more servers and bandwidth, and the vendor passes some of that cost on.

That cost changes the questions you should ask about your tools:

- **How deep does it track now?** A report that shows "not in top 100" may really mean "not in top 20".
- **How often are deep positions refreshed?** Page one and deeper pages may not be checked at the same rate.
- **Did the history break?** A chart that starts at top 50 in October 2025 cannot be compared with one that covered the top 100 before.
- **Where does the data come from?** A licensed data provider and a home-made script that queries Google are not the same thing, legally or in quality.

For most businesses, losing positions 51-100 hurts less than it sounds, because a page at position 70 gets almost no clicks. What you lose is an early warning: a new page climbing from 80 to 45 used to be visible weeks before it reached page one.

## How do you read your Search Console data across September 2025?

Start from clicks, mark the date and compare like with like. A drop in impressions with steady clicks is a measurement change, not lost visitors:

1. **Annotate the dates.** Write "11-12 Sept 2025: num=100 stops" and "27 Apr 2026: impressions logging fix" in your report or dashboard, so no one reads the step as a ranking drop.
2. **Lead with clicks.** Google says the logging error did not affect clicks, and a tool that loads a page does not click.
3. **Split by device.** In the Performance report, filter or compare Desktop and Mobile. The September change showed mainly on desktop.
4. **Use weekly or monthly granularity.** Google's help for the Performance report suggests it for long-term trends, because it smooths out weekends and holidays.
5. **Compare like with like.** Impressions before and after September 2025 were counted differently, so compare clicks year over year and impressions only within one period.
6. **Read average position by query, not for the whole site.** A site-wide average that "improved" from, say, 30 to 18 tells you about the counting change; the position of your top queries tells you about your business. Our guide to [checking your Google ranking accurately](/blog/check-google-ranking-accurately) explains the limits of average position.

For a client or a manager, one sentence usually works: fewer automated views are counted now, so impressions are smaller and closer to what real people see.

## What should you track now?

Track the positions that bring visitors, and get your own data from Google's official source. A practical setup looks like this:

- **Your own site: Search Console.** The Performance report and its API give real clicks, impressions and average position by query, page, country and device. The API returns up to 25,000 rows per request and is the official route for daily exports; we walk through a setup in [how to automate SEO rank tracking](/blog/serp-rank-tracking).
- **Focus on the top 10 and top 20.** That is where clicks happen. Report position bands (1-3, 4-10, 11-20) rather than a single average.
- **Track a short list deeper.** To see a new page climbing from position 60, track a few dozen keywords deep, not thousands.
- **Watch the results page, not just the number.** Ads, map results and AI Overviews can push a "position 3" link below the first screen.
- **Export regularly.** The Performance report keeps a limited window of history, so store your own copy before old data rolls off.

## Why not collect the 100 results yourself?

Because automated searches break Google's rules, whatever the number of pages. Google's [spam policies](https://developers.google.com/search/docs/essentials/spam-policies) describe "machine-generated traffic", meaning automated queries sent to Google, and name scraping results for rank-checking purposes as an example. Google says such activity violates its spam policies and its Terms of Service.

Paging through ten results at a time does not change that, and neither does spreading the queries across many IP addresses. The legitimate options are Search Console for your own site, a licensed rank-tracking or search-data service for competitors and many countries or cities, and manual checks at human scale.

## Where does a location check fit in?

A manual check from the right place answers the one question Search Console cannot: what a searcher in a specific city sees. Search Console groups data by country, not city, and the results in Leeds and London can differ for the same query.

When you are signed out and your browser does not share its location, Google estimates where you are from your IP address. A proxy is a server that passes your traffic on, so the site sees the proxy's IP address instead of yours. With a [Residential Proxy](https://proxynet.io/residential-proxy), that address belongs to a home internet connection in the country and city you choose, so a manual search shows roughly what a local searcher sees. Keep these checks to a handful of searches by a person, or use an approved tool;

our [SEO proxy page](/seo-proxy) describes the country and city targeting for this kind of work.

## Where does the num=100 change show up?

Agencies see the change in monthly client reports, where impression charts show a step in September 2025 that needs a note. In-house SEO teams see it in dashboards that blend Search Console with tool data. Online shops tracking hundreds of product keywords see it in their tool's depth and price. Local businesses notice it less, because their important searches sit on page one. Content teams that used deep positions as an early signal for new articles need a smaller, deeper keyword list instead.

## Common mistakes

- **Treating the impression drop as a penalty.** If clicks did not fall, the site did not lose visitors.
- **Celebrating the average position jump.** The number improved because deep impressions stopped being counted, not because pages climbed.
- **Comparing impressions year over year across September 2025.** The two periods were counted differently. Compare clicks instead.
- **Forgetting the logging error.** Impressions from 13 May 2025 to 27 April 2026 carry a separate reporting issue.
- **Trusting "not in top 100" from a tool that now tracks less.** Check each tool's current depth.
- **Writing a script to fetch ten pages per keyword.** It breaks Google's terms; use Search Console or a licensed service.

## Decision guide

| Need | Recommendation |
|---|---|
| Your own rankings and traffic | Search Console Performance report |
| Daily data for dashboards | Search Console API export |
| Explaining a drop in a report | Clicks first, dates annotated |
| Competitors' positions | A licensed rank-tracking service |
| Early signal for new pages | A small keyword list tracked deep |
| What a searcher in one city sees | Manual check with a city-level IP |
| Year-over-year comparison | Clicks, weekly or monthly |

## Frequently asked questions

### Does num=100 still work on Google?

No. It stopped working around 11-12 September 2025, and Google told Search Engine Land the parameter is not something it formally supports, so do not plan on it coming back.

### Why did my impressions drop in September 2025?

Most likely because impressions caused by tools that loaded 100 results at once stopped being counted. The drop showed mainly on desktop. If your clicks also fell, look for a real cause such as a ranking change or a technical problem.

### Why did my average position improve after the num=100 change?

Because deep impressions at positions such as 60 or 90 left the calculation. Average position is an average over the places where your site appeared, so removing many deep places lowers the number, which reads as better, without any page moving.

### Did Google officially announce the removal of num=100?

No. Google did not announce it in advance. Its comment came a week after the change, when a Google spokesperson told Search Engine Land on 18 September 2025 that "the use of this URL parameter is not something that we formally support".

### How do rank trackers get the top 100 now?

They need about ten requests of ten results each instead of one, which costs roughly ten times more. A tool may therefore track fewer positions, refresh deep ones less often or charge more, so ask your vendor how deep it tracks today.

### Can I still see results for another country without num=100?

Yes. The number of results per page has nothing to do with location, and Google's own region settings or a manual check from the right location still work; we compare the options in [how to search Google from another country](/blog/google-search-different-country).

## Summary

The `num=100` parameter once let Google show 100 results on one page, and rank trackers used it to read the top 100 in one request. It stopped working around 11-12 September 2025, and Google said it does not formally support it. Deep tracking now costs about ten times as much, and many sites saw desktop impressions fall while average position improved, a change in measurement rather than in visitors. Read reports with clicks first and the dates annotated, take your own data from Search Console and its API, track the top 10 and 20 closely and a short list deeper, and check local results by hand from the right city when the country-level view is not enough.
