# Demand Check

> Pull free search-demand signals (Google/YouTube autocomplete + Google Trends) to verify an idea has real demand or to mine rising queries for new ideas. Trigger with /demand-check <idea, keyword(s), or seed space>, or when the user asks "is anyone searching for X", "is X trending", "what's trending around X", or wants demand evidence for an idea-check or a demand score.

- Skill: `davemaynard/demand-check` (Agent Skill, multi-file: 2 files)
- Install (CLI): `npx skillmds@latest add davemaynard/demand-check`
- Raw SKILL.md: https://api.skillmd.com/api/skills/davemaynard/demand-check/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: davemaynard (https://skillmd.com/u/davemaynard)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/davemaynard/demand-check

---


# Demand Check — free search-demand signals

Two jobs, one tool:
1. **Verify** — given an idea or keyword(s), find out whether real people are searching for it, and whether interest is rising, steady, cooling, or below the noise floor.
2. **Ideate** — given a seed space, mine autocomplete long-tails and Trends "rising related queries" for demand that exists but isn't served.

## The tool

Everything runs through one CLI. Paths are relative to this skill's directory (the folder containing this SKILL.md); set `SKILL_DIR` to it first:

```bash
SKILL_DIR=<directory containing this SKILL.md>    # e.g. ~/.claude/skills/demand-check
PY=$SKILL_DIR/.venv/bin/python
CLI=$SKILL_DIR/scripts/demand.py

$PY $CLI suggest "fl studio mcp"                  # autocomplete long-tails (google+youtube), instant, unlimited
$PY $CLI suggest "ableton" --deep                 # + a-z expansion (~27 requests, still instant-ish)
$PY $CLI trend "fl studio mcp"                    # 12-mo interest-over-time, sparkline + verdict
$PY $CLI trend "suno" --timeframe "today 5-y" --geo US
$PY $CLI explore "ableton mcp"                    # trend + RISING/top related queries + regions ← ideation goldmine
$PY $CLI compare "suno" "udio"                    # 2-5 terms; see Known issues
$PY $CLI --json trend "x"                         # raw JSON for any subcommand
```

If the venv is missing (fresh machine): `cd "$SKILL_DIR" && python3 -m venv .venv && .venv/bin/pip install -q trendspyg`.

## Workflow

**Verify mode** (idea/keyword given):
1. `suggest` the core term(s) first — it's free, instant, and unlimited. Autocomplete *presence* is itself a demand signal; a specific repo/brand name appearing in autocomplete means demand with thin supply.
2. `trend` (or `explore`) the 1-3 terms that matter. Read the verdict line; quote the sparkline and peak date in your answer.
3. Translate to a call: RISING/STEADY + rich autocomplete tail = demand verified. BELOW NOISE FLOOR + no autocomplete = no search demand (which kills SEO plays but NOT community/agent-distribution plays — say which applies). Feed the evidence into /idea-check (or whatever scorecard the project keeps) as dated facts.

**Ideate mode** (seed space given):
1. `suggest --deep` on the seed → harvest the long-tail list.
2. `explore` the seed and 2-3 strongest tails → collect **rising related queries** (the "Breakout"/"+N%" rows are unserved demand in motion).
3. Cluster what comes back; present 3-5 candidate demand pockets with their evidence, not a firehose.

## Reading the data honestly

- Trends values are Google's **relative 0-100 index**, scaled to each term's own peak — never absolute volume, and **not comparable across separate pulls**. For absolute volume ranges: Google Ads Keyword Planner (free with an Ads account). For your own sites: Search Console or your analytics.
- Niche terms: `explore` often returns **no related queries** below a volume threshold — that's expected, not an error. The trend line and autocomplete still work at that size.
- A "Breakout" rising query = >5000% growth, usually from near-zero. Exciting but verify the base isn't trivia (song lyrics, a news spike).

## Known issues & rate limits (as of 2026-07)

- **Trends pulls are headless-browser loads, not API calls.** Do a handful per session, spaced out. If a call that worked minutes ago starts throwing `BrowserError` ("did not render the chart"), Google is **soft-blocking** — stop, wait 15-60 min. Don't burn retries; confirmed live 2026-07-22.
- **`compare` (shared-scale) is broken upstream** in trendspyg 1.1.0 — the multi-term page never renders headless. The CLI auto-falls back to per-term pulls and warns that magnitudes aren't comparable. Check for a trendspyg fix occasionally (`pip install --upgrade trendspyg`); upstream: github.com/flack0x/trendspyg/issues.
- `suggest` (autocomplete) has no such limits — lean on it freely.
- Background: pytrends is archived (Apr 2025); Google's official Trends API is still an allowlisted alpha. trendspyg is the maintained free path.

## Answer-engine (AEO) caveat

There is **no free query-level trend data for ChatGPT/Perplexity/Claude usage** — the visibility trackers (Profound, Peec, Otterly) are paid SaaS (skip them for a free pass). Free proxies: the signals above + Reddit/HN mention velocity + GitHub star velocity + AI-referral traffic in your analytics or Search Console.