Research

Research is what turns a website you have connected into a strategy you can act on. It has four moving parts: the brief that says what the business sells, the target that says which market to measure, the run that does the work, and the evidence it leaves behind.

All of it lives under StrategyResearch, in a workspace bar with five panes and three refreshable evidence lanes.

The Research brief

Five fields ground every research run and every AI writer in what the business actually sells. There is no Save button — the caption reads Saves when you leave the field, then Saved.

Two details people miss. What the last draft left out lists rejected terms, services that were derived from a call-to-action, and competitor domains that were parked, each with a reason. And pinned competitors appear here read-only — they are managed in the Evidence pane.

Country and market inputs are not on the brief. They live in SettingsContent & page rules, and the brief links across to them.

Research targets, markets and ceilings

A research target is a domain paired with one country market. Every keyword, gap, SERP and audit surface reads from one. Creating a target is free.

Markets are country-only because the research provider rejects sub-country codes. Targets that have spend history are archived rather than deleted.

Running research

The ten stages of a full run are Intake, Demand, SERP, Entities, Inventory, Baselines, Clusters, Charter, Backlog and Preview. A monthly adjustment runs Inventory, Baselines and an AI review of the active strategy, and does not read the brief.

Run states

State	Badge	What to do
Queued	pending	Nothing — it has not started
Running	running	Watch the stage counter
Awaiting synthesis	awaiting_synthesis	Nothing — it is assembling the draft
Needs you	awaiting_owner	Read the reason chip and resume or cancel
Complete	complete	View draft, then activate from the hero
Failed	failed	Retry stage — it resumes from the evidence already captured
Cancelled	cancelled	Evidence captured is kept; spend already recorded is not refunded

A parked run always names its reason. The three you will meet are Cap reached (raise the cap and resume, or cancel), draft conflict (an open draft already exists), and ungrounded seeds (the brief read as prose, or Search Console returned only brand queries).

The five panes

The workspace bar carries five pane tabs, each with a live badge.

Pane	Badge	What it holds
Brief	needs input when empty or auto-filled	The five brief fields
Run	queued / running / needs you / failed	Scope, cap and run controls
Keywords	row count	The tracked-keyword ledger
Ideas	idea count	The content-ideas radar
Evidence	source count	Everything the AI has read

Panes stay mounted when hidden, so drafts, selections and pager position survive switching. Arriving with no pane in the URL picks one by lifecycle — a brief that needs input first, then a run that needs you. A ?researchView= link always wins, and a run finishing mid-read will not yank you elsewhere.

The three evidence lanes

The same bar carries the three refreshable data streams, each with a freshness pill.

A freshness pill older than 35 days renders as a stale badge. Adding a manual note never freshens the site-knowledge lane — only the crawl itself does.

Evidence: what the AI has read

The Evidence pane is the honest answer to how does it know that? It opens with Evidence at a glance: knowledge sources, retrieval chunks, last pull cost, pinned competitors, structure audit pages and evidence warnings.

• Knowledge sources — crawled pages, uploaded files, transcripts, URLs and notes. Add source attaches one; Delete source removes it, its embeddings and any uploaded file.
• Competitors — pin a domain here and research, and Keyword Gap's defaults, will use it.
• Content structure intelligence — words, H2 headings and internal links as paired bars against competitors, each pair scaled to its own maximum. It says No content-structure audit yet until you run Refresh structure.

Content ideas

The Ideas pane is an opportunity feed: striking-distance keywords from Search Console, competitor top-ranking pages, trending news and RSS. Eight per page, sorted striking-distance first by design.

Tracked keywords

The Keywords pane is the website's working keyword set — volume, difficulty, intent, CPC and rank.

• In report decides which keywords a client ever sees. It is capped, and brand-carrying keywords are never included.
• Add to list… (free, up to 100) and Track positions (free) work without a linked project.
• Measure now buys a fresh rank check for the report set. Owner only.
• Use with Claude opens the brief composer for the selection.

The In report switches are unavailable, rather than guessed, until the server's membership summary loads.

The other research surfaces

Beyond the workspace, the rail carries the market-research surfaces. They read from cached data for free and only spend on an explicit two-click control.

Surface	What it answers	Paid control
Domain Overview	How a domain performs in its market — organic and paid metrics, position and intent charts, top keywords, competitors	Refresh
Keyword Overview	One keyword in depth — volume, difficulty, intent, CPC, twelve-month trend, SERP features	The analysis, and Check live SERP
Keyword Magic	Keyword ideas from a seed, with match types, a questions filter and rich filters	Find keywords, Get more keywords, Search more with these filters
Keyword Gap	What competitors rank for and you do not	Compare

Two rules apply across all of them. A matching set bought in the last 30 days comes back free. And changing a filter, a match type or a device only re-reads what you already bought — filters never spend. Keyword Magic returns up to 500 ideas per run, and 300 on the Exact and Related match types.

Troubleshooting

Symptom	Cause	Fix
Start research is disabled with an amber line about the brief	Neither niche nor primary services is set	Fill them in the Brief pane, or click Draft from the website
The run parked at Cap reached	It hit its cost cap mid-pipeline	Type a higher value in Raise cap $ and Resume, or Cancel
The run parked saying entries read as a sentence	The seeds could not be grounded	Fix the brief so each field is a term, confirm Search Console has data, then Resume
The run parked saying only brand searches were found	Search Console returned brand queries only	Fill the brief's services so research has non-brand seeds
A paid lane is disabled for you	You are a member, not an owner or admin	Ask an owner; free lanes still work
Buttons read Run anyway	The monthly SEO budget is reached	Wait, raise the budget, or click through as an owner to record an acknowledged overrun
Content structure says there is no audit yet	Refresh structure has never run	Run it, or leave it — it is a paid, optional lane
The site-knowledge pill is stale after a crawl	The crawl was truncated, or the site is client-rendered	Retry site crawl; for a client-rendered site, add sources by hand

Related

• The living strategy
• Keyword Gap
• Set up a website for SEO
• Troubleshooting