Poisoned SERP Tracker
The Poisoned SERP Tracker is Semji's daily benchmark of 7 SERP API providers. Every morning, Semji sends each one the same 150 Google searches in 21 markets and counts how often it receives a poisoned SERP, the degraded results page Google serves to a scraper it has detected.
On 16 November 2026, 2 of the 6 providers Semji measured received poisoned SERPs on more than 5% of their checks.
One cell per day, coloured by the share of the provider's checks that returned a poisoned SERP.
- SemjiClean0%0 poisoned SERPs in 278 checksSemji, last 7 days: 7 days at or below 1% poisoned SERPs, 0 days above 1% and up to 5%, 0 days above 5%, 0 days not measured.Semji, last 30 days: 12 days at or below 1% poisoned SERPs, 0 days above 1% and up to 5%, 0 days above 5%, 0 days not measured. Left out of these counts: 17 days of lab history and 1 day measured another way.Semji, last 90 days: 12 days at or below 1% poisoned SERPs, 0 days above 1% and up to 5%, 0 days above 5%, 56 days not measured. Left out of these counts: 21 days of lab history and 1 day measured another way.Days over 5%0 of 70 of 120 of 12
- SerperWatch2.5%7 poisoned SERPs in 278 checksSerper, last 7 days: 0 days at or below 1% poisoned SERPs, 7 days above 1% and up to 5%, 0 days above 5%, 0 days not measured.Serper, last 30 days: 2 days at or below 1% poisoned SERPs, 10 days above 1% and up to 5%, 0 days above 5%, 0 days not measured. Left out of these counts: 17 days of lab history and 1 day measured another way.Serper, last 90 days: 2 days at or below 1% poisoned SERPs, 10 days above 1% and up to 5%, 0 days above 5%, 56 days not measured. Left out of these counts: 21 days of lab history and 1 day measured another way.Days over 5%0 of 70 of 120 of 12
- Bright DataNot measuredBright Data, last 7 days: 0 days at or below 1% poisoned SERPs, 0 days above 1% and up to 5%, 0 days above 5%, 7 days not measured.Bright Data, last 30 days: 0 days at or below 1% poisoned SERPs, 0 days above 1% and up to 5%, 2 days above 5%, 10 days not measured. Left out of these counts: 17 days of lab history and 1 day measured another way.Bright Data, last 90 days: 0 days at or below 1% poisoned SERPs, 0 days above 1% and up to 5%, 2 days above 5%, 66 days not measured. Left out of these counts: 21 days of lab history and 1 day measured another way.Days over 5%Not measured2 of 22 of 2
- OxylabsClean0.36%1 poisoned SERP in 278 checksOxylabs, last 7 days: 6 days at or below 1% poisoned SERPs, 1 day above 1% and up to 5%, 0 days above 5%, 0 days not measured.Oxylabs, last 30 days: 11 days at or below 1% poisoned SERPs, 1 day above 1% and up to 5%, 0 days above 5%, 0 days not measured. Left out of these counts: 17 days of lab history and 1 day measured another way.Oxylabs, last 90 days: 11 days at or below 1% poisoned SERPs, 1 day above 1% and up to 5%, 0 days above 5%, 56 days not measured. Left out of these counts: 21 days of lab history and 1 day measured another way.Days over 5%0 of 70 of 120 of 12
- DataForSEOAlarm36.3%101 poisoned SERPs in 278 checksDataForSEO, last 7 days: 0 days at or below 1% poisoned SERPs, 0 days above 1% and up to 5%, 7 days above 5%, 0 days not measured.DataForSEO, last 30 days: 0 days at or below 1% poisoned SERPs, 0 days above 1% and up to 5%, 12 days above 5%, 0 days not measured. Left out of these counts: 17 days of lab history and 1 day measured another way.DataForSEO, last 90 days: 0 days at or below 1% poisoned SERPs, 0 days above 1% and up to 5%, 12 days above 5%, 56 days not measured. Left out of these counts: 21 days of lab history and 1 day measured another way.Days over 5%7 of 712 of 1212 of 12
- cloroWatch2.2%6 poisoned SERPs in 278 checkscloro, last 7 days: 4 days at or below 1% poisoned SERPs, 3 days above 1% and up to 5%, 0 days above 5%, 0 days not measured.cloro, last 30 days: 6 days at or below 1% poisoned SERPs, 3 days above 1% and up to 5%, 0 days above 5%, 3 days not measured. Left out of these counts: 17 days of lab history and 1 day measured another way.cloro, last 90 days: 6 days at or below 1% poisoned SERPs, 3 days above 1% and up to 5%, 0 days above 5%, 59 days not measured. Left out of these counts: 21 days of lab history and 1 day measured another way.Days over 5%0 of 70 of 90 of 9
- AnakinClean0.74%2 poisoned SERPs in 270 checksAnakin, last 7 days: 6 days at or below 1% poisoned SERPs, 1 day above 1% and up to 5%, 0 days above 5%, 0 days not measured.Anakin, last 30 days: 11 days at or below 1% poisoned SERPs, 1 day above 1% and up to 5%, 0 days above 5%, 0 days not measured. Left out of these counts: 17 days of lab history and 1 day measured another way.Anakin, last 90 days: 11 days at or below 1% poisoned SERPs, 1 day above 1% and up to 5%, 0 days above 5%, 56 days not measured. Left out of these counts: 21 days of lab history and 1 day measured another way.Days over 5%0 of 70 of 120 of 12
- SerpApiAlarm36.7%102 poisoned SERPs in 278 checksSerpApi, last 7 days: 0 days at or below 1% poisoned SERPs, 0 days above 1% and up to 5%, 7 days above 5%, 0 days not measured.SerpApi, last 30 days: 0 days at or below 1% poisoned SERPs, 0 days above 1% and up to 5%, 12 days above 5%, 0 days not measured. Left out of these counts: 17 days of lab history and 1 day measured another way.SerpApi, last 90 days: 0 days at or below 1% poisoned SERPs, 0 days above 1% and up to 5%, 12 days above 5%, 56 days not measured. Left out of these counts: 21 days of lab history and 1 day measured another way.Days over 5%7 of 712 of 1212 of 12
- 1% or less
- Above 1%, up to 5%
- Above 5%
- Not measured
Which SERP API returned the most reliable Google results?
On 16 November 2026, Oxylabs returned the highest share of reliable SERPs among the 6 providers Semji measured: 95% of its checks (264 of 278).
Providers are ranked by share of reliable SERPs; on a tie, the one with fewer poisoned SERPs goes first. Each figure shows the number of checks behind it, and the Semji row sits apart, without a rank.
| Provider | Share of checks by verdict | Agreement | Answered | Median latency | |||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Reliable | Doubtful | Operator ignored | Poisoned | Empty | Failed | Pages in common /10 | First try | Within 5 tries | Standard | AI Overview | |
| 1Oxylabs | Reliable95%264 of 278 | Doubtful3.2%9 of 278 | Operator ignored0.72%2 of 278 | Poisoned0.36%1 of 278 | Empty0%0 of 278 | Failed0.72%2 of 278 | Pages in common /108n = 128 | Answered · First try77.3%n = 278 | Answered · Within 5 tries99.3%n = 278 | Median latency · Standard41.5 sn = 128 | Median latency · AI Overview44.5 sn = 128 |
| 2cloro | Reliable92.8%258 of 278 | Doubtful4.3%12 of 278 | Operator ignored0.72%2 of 278 | Poisoned2.2%6 of 278 | Empty0%0 of 278 | Failed0%0 of 278 | Pages in common /109n = 128 | Answered · First try99.6%n = 278 | Answered · Within 5 tries100%n = 278 | Median latency · Standard14.2 sn = 128 | Median latency · AI Overview20.9 sn = 128 |
| 3Anakin | Reliable89.6%242 of 270 | Doubtful8.5%23 of 270 | Operator ignored0.37%1 of 270 | Poisoned0.74%2 of 270 | Empty0%0 of 270 | Failed0.74%2 of 270 | Pages in common /108n = 124 | Answered · First try99.3%n = 270 | Answered · Within 5 tries99.3%n = 270 | Median latency · Standard12.3 sn = 124 | Median latency · AI Overview11.9 sn = 124 |
| 4Serper | Reliable87.1%242 of 278 | Doubtful5.8%16 of 278 | Operator ignored4.7%13 of 278 | Poisoned2.5%7 of 278 | Empty0%0 of 278 | Failed0%0 of 278 | Pages in common /109n = 128 | Answered · First try100%n = 278 | Answered · Within 5 tries100%n = 278 | Median latency · Standard1.3 sn = 128 | Median latency · AI Overview1.3 sn = 128 |
| 5SerpApi | Reliable36.7%102 of 278 | Doubtful22.7%63 of 278 | Operator ignored4%11 of 278 | Poisoned36.7%102 of 278 | Empty0%0 of 278 | Failed0%0 of 278 | Pages in common /106n = 128 | Answered · First try100%n = 278 | Answered · Within 5 tries100%n = 278 | Median latency · Standard1.2 sn = 128 | Median latency · AI Overview1.3 sn = 128 |
| 6DataForSEO | Reliable17.6%49 of 278 | Doubtful41%114 of 278 | Operator ignored5%14 of 278 | Poisoned36.3%101 of 278 | Empty0%0 of 278 | Failed0%0 of 278 | Pages in common /102n = 128 | Answered · First try97.8%n = 278 | Answered · Within 5 tries100%n = 278 | Median latency · Standard4.8 sn = 128 | Median latency · AI Overview8 sn = 128 |
| Bright Data | Not measured. An account-side problem kept the provider out of this edition: a key refused (HTTP 401, 402 or 403), credits or quota used up, a missing key, the provider switched off, or searches never sent. Semji does not count it as a failure. | ||||||||||
| Semji | Reliable98.9%275 of 278 | Doubtful1.1%3 of 278 | Operator ignored0%0 of 278 | Poisoned0%0 of 278 | Empty0%0 of 278 | Failed0%0 of 278 | Pages in common /109n = 128 | Answered · First try99.6%n = 278 | Answered · Within 5 tries100%n = 278 | Median latency · Standard1.2 sn = 128 | Median latency · AI Overview1.3 sn = 128 |
| Provider | Share of checks by verdict | Agreement | Answered | Median latency | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Reliable | Doubtful | Operator ignored | Poisoned | Empty | Failed | Pages in common /10 | First try | Within 5 tries | Standard | |
| 1Oxylabs | Reliable98.4%126 of 128 | Doubtful0.78%1 of 128 | Operator ignored0%0 of 128 | Poisoned0%0 of 128 | Empty0%0 of 128 | Failed0.78%1 of 128 | Pages in common /108n = 128 | Answered · First try75%n = 128 | Answered · Within 5 tries99.2%n = 128 | Median latency · Standard41.5 sn = 128 |
| 2Serper | Reliable96.1%123 of 128 | Doubtful1.6%2 of 128 | Operator ignored0%0 of 128 | Poisoned2.3%3 of 128 | Empty0%0 of 128 | Failed0%0 of 128 | Pages in common /109n = 128 | Answered · First try100%n = 128 | Answered · Within 5 tries100%n = 128 | Median latency · Standard1.3 sn = 128 |
| 3cloro | Reliable91.4%117 of 128 | Doubtful5.5%7 of 128 | Operator ignored0%0 of 128 | Poisoned3.1%4 of 128 | Empty0%0 of 128 | Failed0%0 of 128 | Pages in common /109n = 128 | Answered · First try100%n = 128 | Answered · Within 5 tries100%n = 128 | Median latency · Standard14.2 sn = 128 |
| 4Anakin | Reliable90.3%112 of 124 | Doubtful8.1%10 of 124 | Operator ignored0%0 of 124 | Poisoned0.81%1 of 124 | Empty0%0 of 124 | Failed0.81%1 of 124 | Pages in common /108n = 124 | Answered · First try99.2%n = 124 | Answered · Within 5 tries99.2%n = 124 | Median latency · Standard12.3 sn = 124 |
| 5SerpApi | Reliable35.2%45 of 128 | Doubtful24.2%31 of 128 | Operator ignored0%0 of 128 | Poisoned40.6%52 of 128 | Empty0%0 of 128 | Failed0%0 of 128 | Pages in common /106n = 128 | Answered · First try100%n = 128 | Answered · Within 5 tries100%n = 128 | Median latency · Standard1.2 sn = 128 |
| 6DataForSEO | Reliable14.8%19 of 128 | Doubtful47.7%61 of 128 | Operator ignored0%0 of 128 | Poisoned37.5%48 of 128 | Empty0%0 of 128 | Failed0%0 of 128 | Pages in common /102n = 128 | Answered · First try97.7%n = 128 | Answered · Within 5 tries100%n = 128 | Median latency · Standard4.8 sn = 128 |
| Bright Data | Not measured. An account-side problem kept the provider out of this edition: a key refused (HTTP 401, 402 or 403), credits or quota used up, a missing key, the provider switched off, or searches never sent. Semji does not count it as a failure. | |||||||||
| Semji | Reliable99.2%127 of 128 | Doubtful0.78%1 of 128 | Operator ignored0%0 of 128 | Poisoned0%0 of 128 | Empty0%0 of 128 | Failed0%0 of 128 | Pages in common /109n = 128 | Answered · First try100%n = 128 | Answered · Within 5 tries100%n = 128 | Median latency · Standard1.2 sn = 128 |
| Provider | Share of checks by verdict | Agreement | Answered | Median latency | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Reliable | Doubtful | Operator ignored | Poisoned | Empty | Failed | Pages in common /10 | First try | Within 5 tries | AI Overview | |
| 1cloro | Reliable94.5%121 of 128 | Doubtful3.9%5 of 128 | Operator ignored0%0 of 128 | Poisoned1.6%2 of 128 | Empty0%0 of 128 | Failed0%0 of 128 | Pages in common /10not available | Answered · First try99.2%n = 128 | Answered · Within 5 tries100%n = 128 | Median latency · AI Overview20.9 sn = 128 |
| 2Oxylabs | Reliable92.2%118 of 128 | Doubtful6.3%8 of 128 | Operator ignored0%0 of 128 | Poisoned0.78%1 of 128 | Empty0%0 of 128 | Failed0.78%1 of 128 | Pages in common /10not available | Answered · First try80.5%n = 128 | Answered · Within 5 tries99.2%n = 128 | Median latency · AI Overview44.5 sn = 128 |
| 3Anakin | Reliable91.1%113 of 124 | Doubtful8.1%10 of 124 | Operator ignored0%0 of 124 | Poisoned0.81%1 of 124 | Empty0%0 of 124 | Failed0%0 of 124 | Pages in common /10not available | Answered · First try100%n = 124 | Answered · Within 5 tries100%n = 124 | Median latency · AI Overview11.9 sn = 124 |
| 4Serper | Reliable90.6%116 of 128 | Doubtful7%9 of 128 | Operator ignored0%0 of 128 | Poisoned2.3%3 of 128 | Empty0%0 of 128 | Failed0%0 of 128 | Pages in common /10not available | Answered · First try100%n = 128 | Answered · Within 5 tries100%n = 128 | Median latency · AI Overview1.3 sn = 128 |
| 5SerpApi | Reliable37.5%48 of 128 | Doubtful25%32 of 128 | Operator ignored0%0 of 128 | Poisoned37.5%48 of 128 | Empty0%0 of 128 | Failed0%0 of 128 | Pages in common /10not available | Answered · First try100%n = 128 | Answered · Within 5 tries100%n = 128 | Median latency · AI Overview1.3 sn = 128 |
| 6DataForSEO | Reliable20.3%26 of 128 | Doubtful39.1%50 of 128 | Operator ignored0%0 of 128 | Poisoned40.6%52 of 128 | Empty0%0 of 128 | Failed0%0 of 128 | Pages in common /10not available | Answered · First try99.2%n = 128 | Answered · Within 5 tries100%n = 128 | Median latency · AI Overview8 sn = 128 |
| Bright Data | Not measured. An account-side problem kept the provider out of this edition: a key refused (HTTP 401, 402 or 403), credits or quota used up, a missing key, the provider switched off, or searches never sent. Semji does not count it as a failure. | |||||||||
| Semji | Reliable99.2%127 of 128 | Doubtful0.78%1 of 128 | Operator ignored0%0 of 128 | Poisoned0%0 of 128 | Empty0%0 of 128 | Failed0%0 of 128 | Pages in common /10not available | Answered · First try100%n = 128 | Answered · Within 5 tries100%n = 128 | Median latency · AI Overview1.3 sn = 128 |
| Provider | Share of checks by verdict | Agreement | Answered | Median latency | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Reliable | Doubtful | Operator ignored | Poisoned | Empty | Failed | Pages in common /10 | First try | Within 5 tries | Operators | |
| 1Oxylabs | Reliable90.9%20 of 22 | Doubtful0%0 of 22 | Operator ignored9.1%2 of 22 | Poisoned0%0 of 22 | Empty0%0 of 22 | Failed0%0 of 22 | Pages in common /10not available | Answered · First try72.7%n = 22 | Answered · Within 5 tries100%n = 22 | Median latency · Operators29 sn = 22 |
| 2cloro | Reliable90.9%20 of 22 | Doubtful0%0 of 22 | Operator ignored9.1%2 of 22 | Poisoned0%0 of 22 | Empty0%0 of 22 | Failed0%0 of 22 | Pages in common /10not available | Answered · First try100%n = 22 | Answered · Within 5 tries100%n = 22 | Median latency · Operators6.3 sn = 22 |
| 3Anakin | Reliable77.3%17 of 22 | Doubtful13.6%3 of 22 | Operator ignored4.5%1 of 22 | Poisoned0%0 of 22 | Empty0%0 of 22 | Failed4.5%1 of 22 | Pages in common /10not available | Answered · First try95.5%n = 22 | Answered · Within 5 tries95.5%n = 22 | Median latency · Operators15.1 sn = 22 |
| 4SerpApi | Reliable40.9%9 of 22 | Doubtful0%0 of 22 | Operator ignored50%11 of 22 | Poisoned9.1%2 of 22 | Empty0%0 of 22 | Failed0%0 of 22 | Pages in common /10not available | Answered · First try100%n = 22 | Answered · Within 5 tries100%n = 22 | Median latency · Operators1.2 sn = 22 |
| 5DataForSEO | Reliable18.2%4 of 22 | Doubtful13.6%3 of 22 | Operator ignored63.6%14 of 22 | Poisoned4.5%1 of 22 | Empty0%0 of 22 | Failed0%0 of 22 | Pages in common /10not available | Answered · First try90.9%n = 22 | Answered · Within 5 tries100%n = 22 | Median latency · Operators4.1 sn = 22 |
| 6Serper | Reliable13.6%3 of 22 | Doubtful22.7%5 of 22 | Operator ignored59.1%13 of 22 | Poisoned4.5%1 of 22 | Empty0%0 of 22 | Failed0%0 of 22 | Pages in common /10not available | Answered · First try100%n = 22 | Answered · Within 5 tries100%n = 22 | Median latency · Operators2 sn = 22 |
| Bright Data | Not measured. An account-side problem kept the provider out of this edition: a key refused (HTTP 401, 402 or 403), credits or quota used up, a missing key, the provider switched off, or searches never sent. Semji does not count it as a failure. | |||||||||
| Semji | Reliable95.5%21 of 22 | Doubtful4.5%1 of 22 | Operator ignored0%0 of 22 | Poisoned0%0 of 22 | Empty0%0 of 22 | Failed0%0 of 22 | Pages in common /10not available | Answered · First try95.5%n = 22 | Answered · Within 5 tries100%n = 22 | Median latency · Operators3.2 sn = 22 |
- Pages in common: the median number of URLs in a provider's top 10 that it shares with the closest other measured provider, on standard SERPs.
- First try: share of checks that returned at least one organic result on the first attempt.
- Within 5 tries: the same share once every attempt has run, with pauses of about 1, 2, 4 and 8 seconds.
- Median latency: how long the successful attempt took, as a median across checks.
How does Semji measure poisoned SERPs?
Every morning, Semji sends one frozen panel of Google searches to each provider and judges every results page by the same rules.
- 01
The same panel for every provider
At 06:00 Paris time, Semji sends each provider the same 150 searches in 21 markets: 128 everyday keywords, asked once for the standard SERP and once for the AI Overview, plus 22 searches using operators such as site: or intitle:.
- 02
Up to five attempts per search
Each search gets up to five attempts, with pauses of 1, 2, 4 and 8 seconds. An answer with no organic result counts as a failed attempt, and Semji logs every attempt with its latency.
- 03
Masked links resolved
Some providers return masked Google links instead of the destination URLs. Semji resolves them before comparing anything, and holds the edition back if fewer than 90% are resolved.
- 04
Rules, then an AI review
Fixed rules give each results page a first verdict, based on the top 10 URLs it shares with other providers, results from another market, duplicates and operators. An AI review then confirms or corrects it.
- 05
Live at 08:00
The edition goes live at 08:00 Paris time if at least five providers were measured and 95% of checks were reviewed. Otherwise the previous edition stays online.
Press kit
About the tracker
The Poisoned SERP Tracker is a daily benchmark published by Semji. Every morning, Semji sends the same 150 Google searches in 21 markets to 7 SERP API providers and reports how often each one receives a poisoned SERP, the degraded results page Google serves to scrapers it has detected. The data is open under CC BY 4.0. Semji is a Generative Visibility Platform: it measures how brands appear in AI search engines and in Google, and helps content teams act on what it finds.
Press contact
For interviews, data questions or corrections, write to contact@semji.com.
Charts
Every chart is stamped with its edition, its number of checks and its licence. Latest links follow the live edition; dated links never change.
Frequently asked questions
What is a poisoned SERP?
A poisoned SERP is a degraded results page Google serves to a scraper it has detected. Semji's Poisoned SERP Tracker flags one when its top 10 shares 3 or fewer URLs with each of the providers that agree with each other, and it also shows results from another market or language, or several pages that would normally rank far down.
Which SERP API providers does the tracker measure?
Semji's Poisoned SERP Tracker measures seven SERP API providers: Serper, Bright Data, Oxylabs, DataForSEO, cloro, Anakin and SerpApi, all sent the same searches each morning. A provider whose account could not be used on a given day is shown as not measured and left out of that day's comparisons, and an edition needs at least five measured providers.
How often is the Poisoned SERP Tracker updated?
Semji publishes a new edition of the Poisoned SERP Tracker every day at 08:00 Paris time, from a full run of the panel that starts at 06:00. A lighter probe also sends three of the searches to every measured provider every two hours, and the page shows when it last ran.
Does Semji use AI to judge the SERPs?
Semji uses an AI model for one step: reviewing the first verdict that fixed rules give each SERP. The model reads each keyword across all providers and confirms or corrects those verdicts, but failed and empty answers keep theirs. Every sentence about the figures is a template filled with the day's numbers.
Does a reliable SERP match what a person sees on Google?
Not always. Semji's Poisoned SERP Tracker has no reference page captured in a browser, so reliable means the top 10 agrees with the other measured providers and the AI review confirms it. If two providers receive the same degraded page, that agreement can mislead, and the review does not always catch it.
How do I report an error in the data?
Write to contact@semji.com. Semji never edits a published edition of the Poisoned SERP Tracker: a correction becomes a new revision with its reason, signed off by two people at Semji and listed in the tracker's corrections log. The series, the CSV file and the charts then use the corrected figures.
Do AI engines cite your brand?
Semji measures how your brand appears in Google and in AI answers, then shows what to publish next. Start with the free GEO audit, or see the platform with our team.
ChatGPTGoogle AI OverviewsPerplexityGeminiClaudeDeepSeek
Serper, Bright Data, Oxylabs, DataForSEO, cloro, Anakin and SerpApi are trademarks of their respective owners. Semji is not affiliated with any of these providers. Each figure describes a provider as Semji measured it on the edition date shown. To report an error, write to contact@semji.com.