Keyword research
How the weekly research agent scores opportunities from your data and turns the best ones into tickets with briefs.
Keyword research runs once a week. The research agent reads your Search Console data, keyword ideas from your enabled sources, your page inventory and your conversions. It scores each opportunity, picks the best ones that are not already tracked, writes a brief for each, and creates tickets on the board.
What the agent reads
| Input | What it adds | Needed? |
|---|---|---|
| Sitemap | The list of pages on your site (the inventory) | Yes, read from your website URL |
| Google Search Console | Queries, pages, clicks, impressions, CTR and position for the last 28 days | Strongly recommended |
| Google Autocomplete | Long-tail searches people really type. No volumes | Free, on by default |
| Bing Webmaster Tools | Related keywords with Bing impressions | Free key, beta |
| DataForSEO | Google search volume, difficulty and intent | Paid, optional |
| Conversions | Pages that appear in converting sessions | Optional, from PostHog or GA4 |
| Business profile and lessons | What you sell, who buys, topics to avoid | Yes |
| Open and recent tickets | Work already planned, so it is not repeated | Automatic |
See Keyword sources to set up each source.
Which Search Console rows it uses
To keep the signal strong, the agent sends only rows that show an opportunity:
- Position 8 to 30 with 20 or more impressions (near page 1).
- More than 200 impressions with CTR under 1% (snippet problems).
- Queries of 3 or more words with 30 or more impressions (long tail).
Seeds
Keyword sources need seed topics. The agent uses Seed topics for keyword research from your business profile. If that is empty, it uses up to 5 items from Products and services.
How opportunities are scored
The agent scores each candidate from 0 to 5 on eight areas, for a total out of 40:
| Area | Question |
|---|---|
| Business fit | Does it match what you sell? |
| Evidence | Is there real data behind it? |
| Demand | Do enough people search for it? |
| Rankability | Can this site rank for it? |
| Conversion path | Does it lead to a conversion page? |
| Urgency | Does it matter now? |
| Effort | How easy is it (5 = easy)? |
| Risk | How safe is it (5 = safe)? |
The agent follows a few rules: first-party data wins over generic suggestions, page-2 rankings with business fit are the fastest wins, long-tail beats broad terms for small sites, and a refresh beats a new page that would compete with an existing one.
Each opportunity gets one type: New article, Refresh, Landing page, Comparison, Internal links or Title/meta.
From opportunity to ticket
- The agent finds twice as many opportunities as you asked for.
- It removes ones that are already tracked, by keyword or by target URL.
- It sorts by priority, then by total score.
- It keeps New tickets per research run (default 5).
- For each one, the brief agent writes a brief. If you set up a SERP source, the brief uses the live top 10 results and "People also ask".
- Tickets are created in To do with source research.
What a brief contains
The brief is the contract for the writer and for QA. It has these sections: job type and action, target keyword cluster, reader and search intent, business reason, SERP summary, title and meta policy, top-of-page structure, outline, product and fact constraints, links and conversion path, assets, schema, and QA acceptance criteria.
You can read and edit the brief on each ticket under Brief (the contract for the writer and QA).
Tip: Changed your product facts, CTAs or business profile? Click Rewrite briefs with current settings on the Ticket board. It rewrites the briefs of all To do, Backlog and Needs human tickets.
Run research by hand
Click Run keyword research on the Overview page. Progress shows on Agent activity.
Read the research report
Go to Insights > Agent reports. Each research report has 3 to 6 summary bullets, the tickets it created, and a Rejected keywords list with the reason for each one. Use rejected keywords to spot topics the agent thinks are off-target. If it is wrong, add a lesson for Research and audit.
Settings
| Setting | Where | Default |
|---|---|---|
| Keyword research day (step 2) | Settings > Weekly schedule | Monday |
| Hour (UTC) | Settings > Weekly schedule | 6 |
| New tickets per research run | Settings > Weekly schedule | 5 (1 to 20) |
| Seed topics for keyword research | Settings > Business profile | Empty |
| Topics to avoid | Settings > Business profile | Empty |
| Target location and Language code | Settings > General | United States, en |
Note: Without Search Console, research still runs on keyword sources and your business profile, but the report lists Search Console as a data gap. Connect it for much better tickets. See Search Console.
Did this page help? Tell us at support@agenticseo.com.