Find the next search opportunity before everyone else.
We read where new AI names appear first — X, Hugging Face, GitHub, sitemaps, Search Console — then map the generic searches behind them. Days or weeks before keyword tools.
Live opportunity feed
Demo data below. Production pages show first-seen dates, sources, target country and evidence links.What is New Term Radar?
New Term Radar (newtermradar.com) is an early keyword opportunity radar for the AI category — a daily pipeline that turns fresh product names into validated, sized search demand.
Every AI wave starts the same way: a name appears where insiders are looking — a lab’s post on X, a model card on Hugging Face, a repository collecting stars at an unreasonable speed. For a few days that name is nearly free to rank for. Then directories file it, comparison sites write it, and the domain squatters register it. New Term Radar exists for exactly that window.
The radar reads the places new AI product names surface first, every single day, and does what no keyword tool can: it maps each new name to the generic, unowned searches behind it — the “ai image detector” behind a detection tool, the “photoshop alternative” behind an editor — sizes that demand, and hands you a decision: prioritize, watch, or skip.
This is the layer Ahrefs, Semrush and Google Trends cannot give you. They are excellent libraries of demand that already exists; they are structurally blind to demand that is being born. A term needs sustained search volume before a keyword database notices it — which usually means weeks after the people who will search it started searching.
Everything on this site is built to be checked. Every lead carries its evidence, its first-seen date and its source. Every lead gets a fourteen-day review against Google Trends, so the site accumulates something no competitor can copy overnight: a measured record of how early these signals actually fire.
How the radar catches emerging AI terms before keyword tools
Six independent sources feed one deduplicated candidate pool. Names are judged by a model, generic demand is sized against Google Ads and Trends, and every conclusion stays traceable to its evidence.
X (Twitter) first mentions
We track the accounts where AI names surface first — research labs, model aggregators and the practitioners who type the name before anyone indexes it. A name two accounts use within 60 days is a lead; a single-account mention needs real engagement to qualify, because capitalized words are often just the start of a sentence.
Hugging Face trending
Models entering the trending top 50 are caught the day they arrive, with quantized copies grouped under the base name so one release does not look like five. Model releases are the fastest-moving signal and the easiest to verify — a number after a name joins it (Opus 5.5 style), which is how buyers actually search.
GitHub breakout repos
A repository under fourteen days old averaging two hundred stars a day — or gaining four hundred since yesterday — is a public, verifiable growth signal. This is the signal behind our documented case: a repo created 2026-09-24 that hit 7,700 stars nine days later, with people searching the founder’s name from day two.
Sitemap surges on watched sites
When several new URLs on a directory or tool site share a word their older pages never carried, something new is being filed under that name. We count surges per distinct URL against each site’s own baseline, so one page with a long slug cannot fake a wave — and we respect every robots.txt we read.
Newly-ranking domains
Domains registered in the last ninety days that already rank for twenty-plus keywords show which names are worth squatting on — and how fast the window closes. This weekly signal is the most expensive in the pipeline, so it runs in batches and its results age into the historical record.
Your own Search Console
The first queries your own site gets impressions for are the earliest demand signal that exists — the market is telling you what it wants before any tool agrees. Connect a read-only Search Console token and those queries join the same candidate pool, deduplicated against everything else.
The keyword table: product names and the generic demand behind them
A new name is half of what you need. The radar reads the evidence once more and writes the generic searches people actually type to do the job — sized, classified, and given a decision.
Product names — a timing play
Names like PhotoCraft spike fast and decay fast. Ranking early for the name itself — a directory page, a comparison, a review — captures the launch window while volume outruns supply. The table estimates daily searches from each name’s own Trends curve and flags the ones still rising.
Generic terms — the durable asset
“ai image detector” is searched ~2,500 times a day and belongs to nobody. The radar proposes these phrases from the product’s evidence, screens how well the name actually answers each phrase, drops the weak matches, and sizes what survives with Google Ads volume plus the term’s own trend shape.
Decisions, not just data
Prioritize: rose recently, still rising, enough daily searches. Watch: rising but small, too early to read, or flat with steady demand. Skip: fading or flat and small. Brand-dependent phrases (like “photoshop alternative”) carry a warning so the trademark call stays yours. Every decision is a versioned rule you can read on the methodology page.
The decision rules, verbatim
Rose in the last 30 days · still up · at least ~300 searches a day
Rising but small · too early to read · or flat with steady demand of that size
Fading · or flat and small — revisit only if the evidence changes
Why speed matters: measured lead time, not promises
The one documented case we can publish end-to-end — signal, timeline, and the review that came back.
Case study: Strata, nine days to 7,700 stars
- 2026-09-24 — the repository is created. Nothing is indexed, no keyword tool lists it.
- 2026-09-25 — the radar reports it as a breakout and starts tracking the search variants people actually type, like “niko strata”.
- 2026-09-26 — those variants show their first impressions in search. No second repo carries the name.
- 2026-10-03 — 7,700 stars. The Google Trends curve for the name finally rises.
- 2026-10-09 — the automatic 14-day review closes: hit, 3 days ahead of the Trends rise. That review pair — signal first, verification later — is the product.
What a lead looks like
Each lead page on this site shows the term, its name score, every piece of evidence with dates and links, the search variants people use, the generic keywords behind it with estimates and decisions, and its review status. If a signal does not have enough evidence to publish, it stays in the internal feed rather than becoming a thin page — we would rather show you less than manufacture volume.
See the Strata lead pageWhat we will not claim
We do not claim every generic term follows its product — in our documented sample most did not move within five days. We do not publish a hit rate until the review sample can support it. And we do not present rule-based scores as probabilities. What we publish instead is the raw material for your own call: evidence, dates, sizes, and decisions you can audit.
Read the methodologyWho it's for
AI tool & PSEO site builders
You need next week’s page topics today. Get the term, the demand behind it and a build-or-skip call.
A typical week: scan the trend board Monday, shortlist three leads with prioritize decisions, spin up directory or comparison pages before the names hit the tools, and re-check the 14-day reviews to see which of last month’s calls hit.
Indie hackers going global
Judge whether a new AI wave has real search demand before you spend a weekend on a domain.
The keyword table answers the question that matters before you register anything: does this new AI thing have a real, sized, unowned search behind it — or is it a model release whose generic phrases are too weak to build on?
SEO consultants & agencies
Deliver early-keyword briefs clients cannot get from Ahrefs or Semrush, with evidence attached.
Every lead exports with its evidence trail — sources, dates, sizes and decisions — so your client sees the reasoning, not just a list. The reports section turns accumulated runs into citable research you can publish under your own brand.
Explore AI keyword opportunities by category
Emerging terms grouped by what the products actually do — the same axis buyers search on.
AI video
Generative video models and tools: Kling, Runway, LTX, Sora-adjacent names and the searches they drive.
BrowseAI image
Image generation and detection — PhotoCraft, SynthID Detector, Qwen Image — and durable terms like “ai image detector”.
BrowseAgents & repos
GitHub breakout repos and agent tooling: the Strata-class signals with the fastest name demand.
BrowseModel releases
Lab releases: strong, fast name signals whose generic phrases need honest screening before you build.
BrowseFrom signal to page: how teams run the workflow
From Monday scan to published page — the loop the product is built around.
Week 1 — discover & validate
Scan the daily board and the weekly digest. For each candidate lead, read the evidence (is it a real product or an announcement?), check the decision (prioritize needs sustained demand, not just noise), and look at the review status of similar past leads before committing time.
Week 2 — build the asset
For a prioritize name: register the domain pattern, publish a directory or comparison page targeting the name and its variants. For a watch keyword: build the better mousetrap page against the generic term — the unowned demand — and let the brand-dependent call follow your own trademark comfort.
Ongoing — measure & compound
Fourteen days after every lead, the review says whether demand materialized. Search Console closes the loop on your own pages: the first impressions are your private confirmation that the signal was real. Hits compound into a library; misses tune the rules. The digest keeps the whole loop on a cadence.
New Term Radar vs. traditional keyword research
| Dimension | Keyword databases | New Term Radar |
|---|---|---|
| Signal lag | Weeks — a term needs sustained volume to enter the database | Same-day: reads first mentions, stars and sitemap surges |
| New-name coverage | None until volume exists | Day one, with name-score and evidence |
| Generic demand | Present, but you must know what to query | Proposed from each product’s evidence, screened and sized |
| Verification | Historical volume only | Automatic 14-day Google Trends review with published hit/miss |
| Decision support | Raw metrics | Prioritize / watch / skip with versioned, auditable rules |
| Best used for | Deep dives on known markets | Finding the next market while it is still cheap to enter |
The two are complements, not substitutes: databases for depth on demand that exists, New Term Radar for the demand being born. That is also why this site publishes its estimates with sources and dates — so the numbers you take into a deep dive are numbers you can defend.
Frequently asked questions
Short answers here; the methodology page carries the detail and the limits.
What is New Term Radar?
New Term Radar is an early-warning system for AI search opportunities. Every day it reads where new AI names appear first — X, Hugging Face trending, GitHub breakout repos, competitor sitemaps and Search Console — then extracts the generic searches behind each name, sizes them, and tells you whether to prioritize, watch or skip.
How is this different from Google Trends or keyword tools?
Keyword tools and Google Trends read step five of a name’s life: the moment search demand is already visible. By then, directories, comparison sites and domain squatters have often acted. New Term Radar reads steps one to four — the first mentions, the star velocity, the sitemap surges — and uses Trends only to verify how early each signal actually was.
What kind of keywords do you surface?
Two kinds. Product names (like PhotoCraft or Claude Haiku 5.5) that are just starting to be searched, and the generic, unowned terms behind them (like “ai image detector”) that people keep typing long after the launch buzz. The second kind is usually the durable opportunity; the first is a timing play.
How accurate are the search volume estimates?
They are estimates, and we label them as such. Volumes come from Google Ads data (which runs roughly two months behind) projected to today along each term’s own Google Trends curve. Every number on this site shows its source and capture date. Terms under a few hundred searches a month are not reliably measurable by anyone — we say so instead of pretending.
Do generic keywords always rise after a new product appears?
No — and we are upfront about that. In our documented October sample, only one of eight generic terms moved within five days of its product appearing. That is exactly why every lead gets a 14-day review against Google Trends, and why hit rates are published with sample sizes instead of promises.
What does “brand-dependent” mean on a keyword?
Phrases like “photoshop alternative” lean on another company’s trademark. The search demand is real, but part of the query belongs to the incumbent brand. We flag these with a warning icon so you can make the trademark call yourself — we never let referral relationships influence scores.
How often is the data updated?
The radar runs daily. Public pages rebuild after each run, the trend board shows each signal’s first-seen date and evidence links, and the weekly digest email summarizes new names, the generic terms behind them, and what the 14-day reviews said.
Who is this for?
People who build pages around new terms: AI tool sites and programmatic SEO operators choosing next week’s page topics, indie developers validating whether a new AI wave has real search demand, and SEO consultants delivering early-keyword briefs their clients cannot pull from Ahrefs or Semrush.
Is the data available in languages other than English?
The interface ships in ten languages — English, 简体中文, Español, Português, 日本語, 한국어, Deutsch, Français, Русский and العربية. Signal data, keywords and evidence stay in English because that is the market the radar tracks: English-language AI search demand.
Start with the public board. Upgrade when it is part of your Monday.
The free tier carries the daily public board, this research section and the weekly digest. Pro unlocks the full scores, competition analysis, watchlists and exports — pricing opens with the first cohort.
The weekly digest
One email a week: the new names, the generic terms behind them, and what the 14-day reviews said. No noise.
Free forever. Unsubscribe anytime.
The weekly digest
One email a week: the new names, the generic terms behind them, and what the 14-day reviews said. No noise.
Free forever. Unsubscribe anytime.