pytrends is an unofficial Python client for the public Trends website. The repo was archived on 17 April 2025. The official Google Trends API is an alpha announced on 24 July 2025. One uses cookies and build_payload. The other uses Google Cloud credentials after an application. They answer the same English question with different numbers. For a hosted JSON path that is neither, see the API reference and pricing.
One is a scrape, one is a Google Cloud product
from pytrends.request import TrendReq
TrendReq(hl="en-US", tz=360).build_payload(["coffee"])
That object never sees trends.googleapis.com. The official client, once approved, is a Google API with a scope and a project. Mixing the two in one class named TrendsClient hides which number is on the chart.
0-100 per request versus consistent scale
Explore, and therefore pytrends, sets the peak in the selected range to 100. A second call for a shorter range rewrites history. The official API's stated point is that values are not rescaled that way, so a caller can append last week without re-pulling five years. That is the whole alpha pitch.
Maintenance: archive versus alpha feedback
pytrends will not merge a cookie fix. The official API asks testers for feedback and may change fields. "Stable" is the wrong word for both. One is frozen. One is unfinished. A production pick is "which failure mode the team can live with."
Trending now is missing from the official alpha pages
trending_searches and realtime methods exist on pytrends because they scrape those HTML surfaces. The official pages talk about 1800 days, daily through yearly buckets, and region or sub-region. They do not list a trending-now method. A newsroom that needs today's board still needs a scrape, a SERP vendor, or another product until Google documents that board on the official API.