Spotting new AI names before Google does

New AI names show up on X, Hugging Face and GitHub weeks before keyword tools list them. new-term-radar is a Python tool that tries to catch them on the day they first appear. It runs every day, pulls names of new models, products, tools and repos from those sources, and turns them into a keyword table. Each row has an estimated search demand and a suggested action. If you write about AI tools, build them or do SEO around them, this is the problem of being too late to a term.

How it works

The radar reads a set of signals and keeps a snapshot of each one. On later runs it reports what changed since the last snapshot. Names from X only count if accounts you list mention them and they are new in the last 60 days. Hugging Face models and Spaces count when they are new to the trending top 50. On GitHub it looks for young repos that are gaining stars fast. On sitemaps it looks for words that several new URLs on a site have in common. There are more signals too: newly registered domains that already rank, Reddit requests for tools, sites that just added their first payment link, revenue growth from TrustMRR, and new queries in your own Search Console.

The part I like comes after that. For each new name it works out the generic terms people would search for to do the same job. It then estimates daily searches, checks the first page of results and marks the row prioritize, watch or skip. After 14 days it checks each lead again against Google Trends. So the tool checks its own guesses.

Try it

git clone https://github.com/zhuyansen/new-term-radar
cd new-term-radar
cp config.example.json config.json
export AISA_API_KEY=...
python -m radar daily --out report.md

The first run only builds snapshots, so on that run you will only see GitHub breakouts. Real diffs start on the second run.

My take

It is standard-library Python, the setup is short, and the prioritize/watch/skip output gives you a decision instead of a list of trends you have to sort yourself. It also ships agent skills and an MCP server (uvx new-term-radar mcp), so you can ask an agent for recent leads. There is an optional GitHub Actions workflow if you want it on a schedule.

The limitations are spelled out in the README, which I like. Most signals depend on one paid AIsa key, billed per call, at roughly $15 a month with the example config. The X signal only knows about the accounts you give it. Search volumes are estimates, and Google Ads data lags by about two months. Nobody has measured hit rates per signal yet, because the 14-day review has only just started. The README also says the new-sites signal was broken as of 2026-10-12. It is tuned for AI tooling, so for any other field you will be editing stop words.

Worth a look if you publish about AI tools or do keyword research in this space and want to know about a name before everyone else has written about it.