Published
Google Removed num=100: What It Changed for Rank Tracking and Search Console
For years, adding &num=100 to a Google search URL returned 100 results on one page. Rank trackers relied on it. One request covered positions 1 to 100 for a keyword.
Around September 13 and 14, 2025, Google stopped honoring the parameter. SISTRIX logged the change on its status page and said it came "without prior notice." Google has not explained it publicly.
The effects reached well beyond rank trackers. Search Console data shifted for many sites, and the cost of deep rank tracking went up by roughly ten times.
What changed
With num=100, a request returned up to 100 organic results. Without it, Google returns one page of about ten results, and reaching position 100 takes ten requests with start offsets of 0, 10, 20, and so on.
BrightLocal described the impact for its own top 50 tracker: a check that used to take one request now takes five. For a top 100 check, it takes ten.
More requests also mean more chances to hit a block, a CAPTCHA, or since 2026 a poisoned page. Each page of a deep check is a separate request that has to succeed, and a single bad page can corrupt the whole check.
Why Search Console impressions dropped
Many site owners saw Search Console impressions fall sharply from the middle of September 2025, while average position improved. SISTRIX's support notes put it plainly: many websites saw "fewer impressions and a better average position," and that "was a result of Google's change, not a loss of rankings."
The usual explanation is that tracker requests had been counted as impressions. A tool loading 100 results registered an impression for every site on positions 11 to 100, even though no person saw them. When those requests stopped, the deep impressions disappeared, and the average position of the remaining impressions looked better. Google hasn't confirmed this, but the timing and the shape of the change fit it.
If you compare Search Console data across September 2025, annotate the change. A drop in impressions from that week is not, by itself, a sign of lost visibility.
How rank trackers responded
Trackers had to choose between charging more, tracking less, or absorbing the cost.
- SISTRIX recorded only the top 10 for most keywords from September 14, 2025, kept 100 positions for its Visibility Index and project keywords, and on September 22 stopped updating desktop SERP data and discontinued its desktop Visibility Index. In May 2026 it began prioritizing crawl depth by search volume and relevance.
- BrightLocal said most customers were unaffected, but moved some accounts on older plans to top 20 tracking by default in October 2025, citing "significant cost increases for fetching data."
Many tools now track the top 10 or 20 by default and charge extra for deeper checks.
How deep should you track now?
Pick a depth for each keyword based on what you do with the answer:
- Top 10 covers almost all clicks and is enough for reporting on keywords where you already rank.
- Top 20 or 30 catches pages close to page one, which is where most optimization work pays off.
- Top 100 is useful for research, new content, and competitor discovery. It rarely needs to run daily.
A common setup checks the top 10 or 20 daily and runs a top 100 sweep weekly or monthly. You can also stop paginating once you find your target domain, since the remaining pages don't change its rank.
What deep tracking costs with a SERP API
When you pay for each request, depth multiplies cost directly. At PrismCrawl's rates of $0.30 to $0.15 per 1,000 successful requests, each page of results is one request:
| Keywords checked daily | Top 10 (1 request) | Top 50 (5 requests) | Top 100 (10 requests) |
|---|---|---|---|
| 1,000 | 30,000 requests a month | 150,000 requests a month | 300,000 requests a month |
| 10,000 | 300,000 requests a month | 1.5 million requests a month | 3 million requests a month |
At $0.15 per 1,000, 3 million requests cost $450. The SERP API cost calculator runs the numbers for your own keywords, locations, and depth. Our rank tracking API comparison shows how other providers price deep checks.
Pagination done safely
If you paginate, a few habits keep deep checks accurate:
- Stop when you find the target. Positions after it don't affect its rank.
- Stop when there is no next page. The PrismCrawl response includes
has_next_page. - Store each page with its request ID, so you can audit one bad page without rerunning the whole check.
- Validate each page. A page with none of the keyword's usual sites may be poisoned. Fetch it again before combining the pages into a ranking.
Our rank tracker build guide has a full example.
Frequently asked questions
When did Google remove the num=100 parameter?
Around September 13 and 14, 2025. SISTRIX logged the change on its status page and said it came without prior notice. Google has not explained it publicly.
Why did my Search Console impressions drop in September 2025?
The most common explanation is that rank tracker requests loading 100 results had been counted as impressions for sites in positions 11 to 100. When num=100 stopped working, those impressions disappeared. SISTRIX attributes the lower impressions and better average position to Google's change itself.
How many requests does a top 100 rank check take now?
About ten. Google returns roughly ten organic results per page, so reaching position 100 requires ten paginated requests using start offsets of 0, 10, 20, and so on up to 90.
Can I still get 100 Google results in one request?
Not from Google directly. Google no longer honors num=100, so 100 results take about ten paginated requests. Some APIs make those requests for you, but each page still has to be fetched.