Technical note3 min read
Why a Google SERP API now returns five results, not ten
Google now unlocks each result link a little later than the one above it, the tenth about 24 seconds after the page. Five results arrive before the long waits: 2,273 of 2,500 live searches answered, 12 at a time, at a median of 4.19 seconds.
2,273 of 2,500 live Google searches answered, 12 at a time, at a median of 4.19 s.
The run's searches in the order sent, a column to every 25, five results asked of each; under them, how long the answers and the refusals took.
- All five results, 2,256
- Fewer, 17
- Refused, 227
Where these figures come from. Every figure about the runs is Vitna's own measurement, published whole: each search's query, result count and time are in the run file. How Google's result links unlock is from Vitna's SERP reference, measured on 3 November 2026.13
2,273 of 2,500 live Google searches came back, at a median of 4.19 seconds. That was the run of 4 November 2026: 2,500 different technical questions, 12 at a time, through the retrieval behind Vitna's POST /v1/serp at its default of five results. Two runs the night before, three at a time, answered 296 of 300. At ten results a search, the median took 26.1 seconds.12 Most of the difference is when Google lets each link on its page be read.
Vitna's runs, times over every search, refusals included.12 The ten-result runs were made under a different setup, so only their times compare.
Google's links now wait their turn
Google's results page no longer links straight to the sites it lists. Each result points at a google.com/goto address carrying an opaque token, and the page holds no publisher address anywhere, in its HTML or its scripts. Asking that address where it leads answers at once for the first two results. Each result further down unlocks later than the one above it, and one asked too early answers with a “Verifying your request” page and a wait instead.3
| Result on the page | Its link unlocks |
|---|---|
| 1 and 2 | At once |
| 3 | About 0.5 s |
| 4 | About 1.05 s |
| 5 | About 2.3 s |
| 6 | About 7.6 s |
| 7 to 10 | About 24 s |
Vitna's measurements on 49 result pages, 3 November 2026, timed from the page settling.3 They are Google's to change.
Vitna returns the publisher's own address for every result, never a /goto link, so it asks Google for each one. That makes a results page exactly as fast as its slowest link.
Why five, not ten
Ten results means waiting for the tenth link, about 24 seconds after the page. Five means waiting about 2.3 seconds. So on 3 November the API's default fell from ten results to five. maxResults still asks for anything from one to ten, and a caller who needs the whole page can have it at the whole page's cost.6 Asking for one or two skips the wait altogether.
Of the 2,500 searches, 2,209 finished in under seven seconds. The answers took a median of 4.16 seconds, and 95 in 100 of them came back within 5.35 seconds.1
What came back short
Of the 2,500, 2,256 came back with all five results. Another 16 lost a result each to doorway pages, which hacked sites publish for whatever people search and which the API leaves out unless asked to keep them.4 One came back with four results, for a reason the run did not record.
The other 227 were refused, the first one sent among them: 222 with serp_blocked and 5 with serp_timeout. The API answers each as a 503 error, never as an empty list, so a caller can tell a refusal from a page with nothing on it. A refusal took longer than an answer, a median of 14.9 seconds.5
The refusals were not spread evenly. Of the first 250 searches sent, 184 were answered; of the 2,250 after them, 2,089 were. The run did not record why.1
The two runs the night before, with three searches in flight, took a median of 3.92 seconds and were refused four times in 300. Across all three runs, 2,569 of 2,800 answered, 90.7 to 92.7% at 95% confidence.12
What the times leave out
Each time runs from the call to the parsed answer on Vitna's search host. The runs were made there, in a process running the deployed code with the service's own configuration, so a caller's trip across the internet and the API's authentication and quotas are not in them.1
Every question was technical: configuration keys, flags, command lines and named techniques. Three runs on one host within a day describe those runs, not all of Google at every hour, and Bing was not measured. The defaults, the limits and every error code are in the reference.6
Sources
- 1DataLive Google SERP runs, 3 and 4 November 2026Vitna, every search with its query, result count and timeread
- 2DocsMeasured deliveryVitna, SERP referenceread
- 3DocsRun on owned infrastructure: Google's /goto links and their holdsVitna, SERP referenceread
- 4DocsDoorway pagesVitna, SERP referenceread
- 5DocsAuthentication, quotas, and failuresVitna, SERP referenceread
- 6DocsRequestVitna, SERP referenceread
More posts
Claude API pricing: how Opus fell from $75 to $20
A million tokens of Opus output cost $75 in May 2025. Opus 5.5 launched at $20 in September. Every Claude model's price today, every change since, and what one coding session costs.
Who owns the AI companies, and who still holds the votes
Elon Musk owns under half of SpaceX, home of xAI, and casts 82% of its votes. Microsoft owns about a quarter of OpenAI and names none of its directors. How the money and the votes came apart at 13 AI labs.
How Beam 1.0 got past its tokens-per-second ceiling
Beam 1.0 runs at 348.8 tokens a second across its hosts and 537.3 on the fastest. One token a pass stops near 300. What sets that ceiling, and the ways past it.
How DeepSeek went from price war to price rise
DeepSeek's cheapest model costs $1.20 per million output tokens at peak, up from $0.28 in April. Every seller's price today, and every release's price and score.
How a stock trader came to own DeepSeek
Liang Wenfeng still controls DeepSeek. His hedge fund paid for it for three years, and the outside investors who bought in this June hold under 9%.
