I'm a data scientist and a professional data engineer, and an MSc student in Computer Science and Engineering, where the courses I chose are about data and AI. I started working with Google Trends when pytrends, the popular Python library for it, was archived in April 2025.
Something didn't match
Then I noticed something was wrong with my tool: what I saw in my normal browser and what my tool got didn't match. The rising related queries for "python" in my tool were things like "walmart near me". I was surprised, and curious.
So on 7 November 2026 I compared one search, "python", United States, past 12 months, panel by panel: what the Google Trends website showed in my browser, and what Google sent to my automated request.
What I found
| Panel | Website, in my browser | Automated request |
|---|---|---|
| Interest over time | 53 weeks | identical, 53 of 53 |
| Interest by state | 51 states | identical, 51 of 51 |
| Interest by city | 34 cities | 3 cities |
| Rising related queries | "python docs", "python 3.15"… | "walmart near me", "lidl near me", "cheap flights 2024", "how to cook pasta"… |
| Related topics | 38 topics | none |
Interest over time and interest by state were identical for both. But the website showed 34 cities in the browser, while the request got only 3. I mean, are you kidding me?
What surprised me more was the rising related queries. They were totally different. The website showed "python docs", while the request got "walmart near me", "lidl near me", "cheap flights 2024" and "how to cook pasta". Only 2 of 25 were the same.
And related topics? 38 on the website. Guess how many the request got. Yes: zero. :D
Why?
I think Google does this by labelling our requests: it treats a real browser as a normal user, and our automated request as a scraper. The requests labelled as scrapers got the decoys.
Why Google does this instead of sending an error is what confuses me. Why, Google? Why send decoy data?
Who should worry
People who use rising queries for content ideas, SEO, keyword research or research papers should worry about this. The decoys look normal, and nothing tells the user that they are fake.
For example, someone finds "baby diapers" as a rising keyword and starts making baby-diaper videos, while the trend he's chasing doesn't even exist.
Check your own tool in one minute
You can really check this in just one minute:
- In your tool, get the rising related queries for "python", United States, past 12 months.
- Open the same search on trends.google.com in your normal browser.
- Compare them.
If the keywords don't match, then boom: you've been fooled by Google, and so has the tool you were using.
What to do instead
I think what people should do is use only the panels that actually match. For something like related queries, use the download button on the Google Trends website itself. And once a month, compare that download with your tool.
Full disclosure
After discovering the issue, I fixed my Google Trends scraper. It now returns only the panels that actually match the website, and leaves out related queries, related topics and cities. I check it myself against the website every month, so I don't get the fake data.
Let me know: is your tool sending decoys too?
Top comments (0)