Google Trends still has no API in 2026. Here's what I use instead

Google announced a Trends API alpha in July 2025. It's now August 2026 and access
is still application-gated. I applied, waited, heard nothing, and eventually built
my own thing.

If you just need trends data in a pipeline or a spreadsheet, here's the honest
lay of the land.

The official alpha is worth applying for anyway — official beats scraped every
time, and if your application goes through and the quotas fit, stop reading here.
Mine didn't go through.

The commercial SERP APIs (DataForSEO, SerpAPI) all have Trends endpoints. They
work. They also start around $50-75/month, which is a strange amount of money to
pay when you want to check thirty keywords once a week. If you're doing enterprise
volume, fine, that's your answer.

I wasn't doing enterprise volume. So I built
Google Trends Explorer on
Apify — you pass keywords, a region and a time range, and get one JSON item per
keyword back:

{
    "keyword": "notion templates",
    "geo": "worldwide",
    "timeRange": "today 12-m",
    "latestValue": 33,
    "averageValue": 68.6,
    "peakDate": "May 3 - 9, 2026",
    "interestOverTime": [{ "date": "Aug 2025", "value": 48 }],
    "relatedQueries": {
        "top": [{ "query": "templates for notion", "value": 100 }],
        "rising": [{ "query": "notion marketplace", "change": "+250%" }]
    }
}

The latestValue / averageValue / peakDate fields are there because half the
time that's all you actually wanted — feeding a 54-point series to an LLM so it
can tell you "interest peaked in May" is a waste of tokens.

Some things you should know before trusting any Trends data, from any source:
the values are Google's sampled, relative 0-100 index within whatever range and
region you picked. Run the same query twice and small numbers can wiggle. It's a
signal for "is this topic growing or dying", not a measurement.

And Google rate-limits these endpoints hard. My actor rotates proxy sessions and
backs off on 429s, but I made one design decision I'd defend anywhere: if every
keyword fails, the run exits FAILED. Loudly. I've been burned by scrapers that
return an empty array and call it success — you find out three weeks later that
your "data pipeline" has been shipping nothing.

Full disclosure: I build and maintain this with AI assistance, on my own account,
and it only touches public aggregate data. If the official API ever goes GA with
sane quotas, I'll be the first to switch.