From a topic to a publish-ready brief.
Point Research Studio at a topic and five agents fan out in parallel — a keyword pool, the top-20 SERP with real headings and People-Also-Ask, an AI-citation probe, brand-fit opportunities, and an evidence extractor. Watch the research land live, keep what's useful, and generate a 12-section brief your writer can ship.
Credit card required · Cancel anytime · 7-day full access
Deep research is three hours of tabs — and none of it survives into the draft.
The old way: open forty tabs, skim ten competitors, paste headings into a doc, hunt for a statistic you half-remember, and write the brief by hand. Hours per topic — and when someone asks where a number came from, nobody can answer.
- One run replaces the afternoon of tabs
- Every source stays on the session, traceable
- It ends in a brief a writer can start from
Five specialist agents fan out at once.
You give it a topic, a goal and an audience. Keywords, SERP, AI Visibility, Opportunities and Evidence each answer a different question about the topic — in parallel, on the same run. Run sequentially you'd get five reports; run together you get one judgement.
- Each lane answers a different question
- Watch research land live — keep or discard
- A sixth agent writes the brief, after the gate
One topic becomes five search phrasings.
Discovery doesn't fire a single query. It derives up to five variants — literal, geographic, commercial, pricing, question. It used to issue exactly one: "Software development services" came back as Glassdoor, Indeed and ZipRecruiter. One job-flavoured phrasing decided the entire evidence base.
- Up to five phrasings, derived automatically
- Every query recorded on the session — strange results are traceable
- No single wording can decide your evidence base again
Six phases that never round up.
Discovering, classifying, scoring, deep search, enriching, ready — with a live counter the whole way. And when a run only partly finishes, the bar says so: 4 of 6 phases finished. Never a full bar on a failed run.
- Live counter: sources found, sources kept, time elapsed
- "Finished with problems" — what completed is safe to use
- A genuinely stuck run says so, with a restart
58 found. 48 kept. Every cut explained.
Deterministic keep-rules score every source as it lands. Weak ones aren't silently dropped — they're grouped under Low quality with a named reason each. And you can discard anything mid-run; it never reaches your brief.
- Keep-rules score and select the strongest sources
- Named reasons: content farm, document viewer, site-builder page…
- "Not shown: 9 discarded." — the count stays visible
A keyword table that tells you what it doesn't know.
Opportunity is scored from five named parts — volume, attainability, value, gap, search-console. Your own Search Console rank sits in the row. And when a metric is missing, it says No data with a reason — never a dash you have to interpret.
- "Not scored" when parts are unknown — never a fake zero
- #7 or "Not ranking" — your real rank, from Search Console
- Volume, difficulty and CPC at your own market
| Keyword | Opportunity | MSV | KD | CPC | Coverage | Your rank | Intent | Pin |
|---|---|---|---|---|---|---|---|---|
| how much does custom software cost Search dataRanking | 720 | 28 | $3.90 | 2 pages | #7 | Question | ★ | |
| custom app development cost Search dataRanking | 480 | 34 | $4.60 | 1 page | #9 | Commercial | ★ | |
| custom software vendor pricing AI-suggested | Not scored | No data ⓘ | 38 | No data | — | Not ranking | Commercial | ★ |
| software company pricing tiers Search data | 590 | 41 | $5.80 | 3 pages | Not ranking | Transactional | ★ | |
| custom software quote template Search data | 320 | 22 | $2.10 | 1 page | #18 | Transactional | ★ |
Demoted keywords say why. Nothing is deleted.
A keyword that quietly vanishes is a defect; one that explains its demotion is a feature. Off-goal marks a keyword that's about your topic but serves a different intent than your goal — kept and pinnable, never auto-recommended. Filtered (6) opens a drawer where every hidden keyword names its reason.
- Off-goal: on-topic, wrong intent — still pinnable by hand
- Filtered reasons: competitor brand, near-duplicate, quality check
- Everything stays on the session; nothing disappears
The pages that actually rank, heading by heading.
Ranked results with their real outlines scraped from the live page, plus the real People-Also-Ask set. AI-topped-up questions are badged Suggested so you always know scraped from generated — and an off-topic question is demoted, never hidden, because it's real search data.
- Real headings from real pages, expandable per result
- "Suggested" badge on anything AI added
- A blocked page says "Couldn't read this page"
Every engine reports its true outcome.
Has data, Found nothing, Didn't complete, Not run — a failed engine never shows a fabricated zero. And when a lane fails, you retry that lane for 1 credit instead of re-running the whole research for 5.
- Citation presence is checked across the same AI answer engines as the AI Visibility tracker
- Who gets cited, who gets named — per engine
- Partial failure is labelled, and what ran stays usable
- Retry one lane: 1 credit, not a fresh 5-credit run
The gaps competitors left open — with the reasoning attached.
Uncovered topics and underserved audiences, scored for impact against your brand. Each card explains itself, so "why this angle?" is answered before the writer asks.
- High / Medium impact, scored against your brand
- Uncovered Topic or Underserved Audience — typed, filterable
- Reasoning on the card, ready to carry into the brief
Statistics with their sources still attached.
The Evidence lane extracts stats, quotes and facts out of the pages that already rank — each one carrying the page it came from. Check a number before you write, not after someone questions it.
- Stat / Evidence / Quote — typed and pinnable
- Source attribution on every card
- Discarded evidence never reaches your brief
Nothing gets written from thin research.
Four buckets must clear a threshold before Generate Brief unlocks — thresholds owned by the server, not hardcoded in the page. The disabled button names what's missing. A lane that genuinely found nothing relaxes to optional instead of deadlocking you. And on a finished session, you can still generate anyway.
- "Add more: Keywords and Evidence" — the button names the bucket
- "(none found — optional)" — no mystery ticks
- "Generate anyway" when you'd rather ship with the pins you have
The 12-section brief: grounded, versioned, sourced.
A full editorial document — outline informed by the pages that actually rank, evidence attached section by section, audience and tone from your brand profile. The SERP-grounded badge means the outline came from live competitor pages; the rank line is computed from your pins, never written by the model.
- Rank line computed from pins, not claimed
- Do's & Don'ts a writer can actually follow
- One click to a pre-filled draft
The model writes it. Code proves it.
After the brief is written, five deterministic validators run over it — code, not a prompt instruction. Any figure not backed by your own pinned evidence is excised: the span, not the bullet. A brief once invented "$8K–$20K" while the session's kept evidence said $40K–$500K. Now that number can't survive.
- Numbers ground only against human-pinned evidence — never the model's own
- Internal vocabulary scrubbed from editorial guidance
- Sources Index rebuilt from the pins, not trusted from the model
- Every section is schema-validated on generation, so nothing comes back half-formed
Research Studio, answered.
Research Studio runs the keyword, SERP, citation and evidence lenses on your topic in one pass, clears each bucket against a threshold, then generates a 12-section brief that hands straight to the Content Editor as a pre-filled draft — from topic to sourced brief in one workflow.
A brief worth writing from settles four things before anyone drafts: the query and the intent behind it, what the pages currently ranking already cover, the angle that adds something they do not, and the evidence backing each claim. Assembled by hand that is a few hours per topic. Research Studio produces the same shape as a 12-section brief with its sources attached, so the writer starts from a position rather than a blank page.
Keyword demand, SERP structure, citations, evidence and competitor coverage each answer a different question about a topic, and running them together is what lets the brief reconcile them — a term with strong volume but no citable evidence behind it is a different proposition from one with both. Run sequentially you would get five reports; run together you get one judgement.
Each research bucket has to clear a threshold before a brief is generated. If a topic cannot gather enough credible sources or evidence, it does not quietly produce a thin brief — it tells you the topic is not ready and which lens came up short. A lane that genuinely surfaced nothing relaxes to optional rather than blocking you forever.
Live search results, the pages currently ranking for the topic, and the sources those pages cite — gathered at run time rather than recalled from a model’s memory. Every claim that reaches the brief carries the source it came from, and any figure that is not backed by your own kept evidence is removed before you see it.
Yes. Research runs against the site you have selected, so the keyword set, the competitor context and the brand voice applied to the brief are that client’s, not a house default. Switching sites switches all of it.
Every paid plan. Starter includes 5 research runs a month, Pro 15 and Agency 30. A trial includes one run — enough to take a real topic all the way through to a brief. See the pricing page for what each plan costs.
Its credits are returned automatically — the charge and its reversal are computed by the same code, so they cannot disagree about the amount. If only one lane failed you can retry that lane on its own rather than paying for the whole run again, and whatever did finish stays usable in the meantime.
Turn your next topic into a brief your team can publish
7-day free trial · credit card required · cancel anytime