Search Demand Reader

Search Demand Reader

Keyword tools estimate. Search Console counts. The counts are limited to searches where one of your pages appeared, and that limit is the point: every query in the export is one you can already win more of, because Google has already shown you for it. This engine reads that export the way a content lead would with a free afternoon: which intents your traffic serves, which topics carry the demand, where you rank past page one for something people search a lot, and where two of your own pages are fighting over one query.

When to use

Before planning content for a quarter, when deciding between a new page and a refresh, when the founder asks "what do people search for", and when branded traffic seems to be all there is.

Not for a site with no Search Console history, and not for sizing a market: a query you never appeared for is not in this data and the engine will not guess it. For the demand you cannot see, use interviews, review-mining, and the sizing method in market-sizing.

Before you start

1. Read the brand/product context first (.agents/product-marketing.md) for the brand words, rival names and the commercial topics that matter.
2. Export Queries for at least three months. Performance > Search results > Queries, all rows. Add the Page dimension if the interface lets you export query-and-page pairs; otherwise export Queries and, separately, filter by page for the few topics you care about.
3. List your brand words (name, misspellings, product names) and the rival brands buyers compare you with.

Method

Intent is read from modifier words, in order: a brand word makes it branded; a rival's name makes it rival-brand; "demo", "trial", "buy", "pricing" and the like make it transactional; "best", "vs", "alternative", "software", "tool" make it commercial; "how", "what", "guide", "template" make it informational; anything else is unqualified, which means the words did not say, not that the engine guessed.

Topics are groups of queries sharing their most common content word, after light stemming. This is coarse on purpose: a word-level grouping you can check beats a clustering you cannot. Read the top queries under each topic before acting on it.

Gaps you already appear for. A topic with 200+ impressions (engine rule) whose impression-weighted average position is worse than 10 (engine rule). Google shows you for it and nobody clicks because you are on page two. These are the cheapest content wins on the list, because the search intent is proven and the page often exists.

Cannibalisation. A query where two or more of your pages each take 20% or more of its impressions (engine rule). Google is choosing between your pages, and the ranking suffers for it.

Branded share of clicks says how much of your organic traffic is people who already knew you. There is no right number; a rising non-branded share is the sign the content programme is working.

Rival-brand searches you appear in are comparison intent you have already earned a place in; comparison-pages is the next step.

Workflow

Step 1 - Export, run, read the verdict.

Step 2 - Take the largest gap. Look at its top queries. If one page already holds them, refresh it with seo-content-brief aimed at that query set. If no page is built for the topic, brief a new one. Do not build a page per query.

Step 3 - Resolve cannibalisation by choosing the page that should win and pointing the other at it (a link, a consolidation, or a canonical), then watch the query's position next quarter.

Step 4 - Put the rival-brand queries into comparison-pages.

Step 5 - Put the rest in the calendar (content-calendar-planning), gaps before new ideas.

Step 6 - Re-run quarterly and track the non-branded share of clicks and the number of gaps.

Run the tool

node resources/search-demand.js --demo
node resources/search-demand.js '{"brand":["acme"],"queries":[{"query":"crm for dentists","clicks":3,"impressions":400,"position":12.1,"page":"/blog/dental-crm"}]}'
node resources/search-demand.js --help

run({ brand, rivals, queries }) returns totals, intents, topics, gaps, cannibalisation, rival-brand queries, findings and a verdict. Deterministic, offline, no dependencies.

Present the result

1. The verdict and the branded share of clicks.
2. Intent table: queries, impressions and clicks by intent.
3. Gaps, each with its top queries and the page that currently appears.
4. Cannibalised queries with the page split.
5. Rival-brand searches you appear in.
6. Topics table (top 15) for the broader picture.
7. Findings, especially missing pages or positions, which limit what can be read.

Guardrails & common mistakes

• Never present an impression count as search volume. It is the number of times your page was shown, which is at most the volume and usually far less.
• Never fill the gap you cannot see. If a topic is absent from the export, say that the data cannot speak to it rather than estimate.
• Unqualified is a category, not a guess. Read those queries by hand before assigning intent.
• One page per topic, not per query. The gap list groups queries for a reason.
• Branded share has no benchmark. Compare it with your own last quarter.
• Three months minimum. A month of queries is mostly noise for a small site.

Related AAJ resources

• Briefing the page that closes a gap: https://aajconsult.com/tools/content-brief-builder
• Sizing the market from accounts rather than searches: https://aajconsult.com/tools/bottom-up-market-sizer
• Where search data fits among other sources: https://aajconsult.com/playbooks/market-research-playbook
• The content system this feeds: https://aajconsult.com/playbooks/content-creation-playbook

Related skills

• content-decay - the same export, read page by page over time
• seo-content-brief - the brief for the gap page
• content-calendar-planning - scheduling gaps and refreshes
• market-sizing - demand you cannot see in search data
• aaj-programmatic-seo - when a gap is a template set, not a page
• review-mining - the language of the demand, from reviews