Auto Research turns a short description of your website into a finished keyword plan. You write a few sentences about what the site is, who it is for and what you intend to publish; ClearSERP works out the seed keywords and competitors, harvests every keyword it can find, ranks each one by how well it fits your brief, and groups the good ones into the articles you should write. The result is a report with four tabs: Articles, Map, Topics and Keywords.
It replaces the manual loop of running Keyword Ideas and Niche Keywords seed by seed, merging the spreadsheets and sorting thousands of rows by hand. Every keyword in the report can still be sent into full SERP analysis or saved to a collection, exactly like a keyword from any other research mode.
How it works
A run has three stages, and you stay in control at the boundary between each one.
- Describe. Write the brief, pick a location and language, choose a depth and optionally add your own domain and constraints. Nothing is charged yet.
- Review the plan. ClearSERP proposes seed keywords, competitor sites and topic angles, and says what it will leave out and how it will rank results. Edit anything, or ask for changes in plain English. Approving the plan charges the depth you chose and starts the run.
- Read the report. The run takes a few minutes to a few tens of minutes depending on depth. You can leave the page; the report is waiting under Auto Research history when it finishes.
Writing a good brief
The brief is the single biggest lever on result quality, because relevance is judged against it. A useful brief answers four questions:
- What the site is about. The subject, in the words your readers use.
- Who it is for. Homeowners or contractors, beginners or professionals, buyers or researchers. Audience decides which intents count as core.
- What you plan to publish or build. Reviews, how-to guides, comparison pages, a tool, a product catalogue. This decides which keyword shapes belong in the article plan.
- What is not relevant. Say it explicitly. "We do not sell wholesale", "not for commercial kitchens" or "no competitor brand names" saves you from a report full of keywords you cannot use.
Adding your own domain is optional. When you give one, ClearSERP uses it to understand what you already cover and to find the competitors that actually overlap with you.
Choosing a depth
Depth controls how many seeds and competitors the run harvests from, and therefore how many keywords it can find. Each depth has a price, shown on the plan before you approve. The price is a maximum: it is charged up front and settled when the run finishes, so a niche with less ground to cover than the depth's ceiling gets the unspent share refunded automatically.
| Depth | Credits per run | Redo articles | Rewrite titles | When to use it |
|---|---|---|---|---|
| Quick | 500 | 200 | 40 | Fewer seed keywords and competitors. Good for a first look. |
| Thorough | 2,000 | 600 | 100 | A wide sweep across seed keywords and competitors. Most people want this. |
| Exhaustive | 4,000 | 1,000 | 200 | The widest sweep. For large topics with a lot of ground to cover. |
Thorough is the right default for most sites. Quick is for checking whether a niche is worth a deeper run; Exhaustive is for broad topics where a Thorough run visibly leaves ground uncovered.
Reviewing the plan
The plan page is where you steer the run before spending anything. It has these sections:
- What you asked for and what we understood: your brief, and the planner's reading of what you do, who your audience is and what you are producing. If the reading is wrong, fix it here; everything downstream depends on it.
- Topic angles: the axes the research will explore. Switch off any angle you do not want covered.
- Seed keywords: the starting points for keyword harvesting. Seeds are sources, not targets; a seed that itself lands low in the report has still done its job if it pulled in good keywords.
- Competitor sites: domains whose ranking keywords and pages will be mined. Remove the ones that are not really your competitors and add any that are missing.
- What we will leave out and how we rank the results: the exclusions and the relevance rubric the run will apply. This is the place to check that your "not relevant" notes were understood.
- Ask for changes: type an instruction such as "focus on beginners" or "drop anything about commercial use" and the planner revises the plan in place.
You can also duplicate an existing plan to start a new run from the same brief, which is the quickest way to try a different depth or a tightened set of constraints.
Constraints
Optional filters narrow what reaches the report: minimum and maximum search volume, keyword difficulty and CPC, a "must contain" and "must not contain" word list, and a questions-only switch.
What the run does
While a run is in progress the report page shows each phase as it completes. In order, the run is:
- Finding keywords from every seed and competitor in the plan
- Removing duplicates
- Reading competitor pages and live search results
- Mapping the subjects of the niche
- Thinking up keywords the research missed, then checking real demand for them
- Checking how the keywords are worded
- Ranking every keyword by relevance and double-checking the best ones
- Grouping keywords into articles and checking real search results for each
- Reviewing the whole plan for duplicates and drift
- Naming topics, writing titles and building the topical map
Two of those phases deserve a note. Keyword harvesting is not limited to the suggestion databases: the run reads competitor pages and live search results, maps the subjects of the niche and originates keywords the databases missed, then checks real demand for them. And search volume is treated as a signal rather than a gate, so a highly relevant keyword with a low reported volume is kept rather than dropped.
Relevance bands
Every keyword is placed in one of four bands, judged against your brief.
| Band | Meaning |
|---|---|
| Core | Exactly what you asked for: the right subject, the right audience, the right kind of content. Articles are built from these. |
| Peripheral | The right audience but a weak fit for what you plan to publish. Left out of the article plan by default; a toggle on the Articles tab adds them. |
| Tangential | Related to the niche but not to your brief. Kept in the Keywords tab for reference, never grouped into articles. |
| Junk | Wrong subject, wrong intent or unusable. Discarded, apart from any keyword you explicitly put in the plan. |
The Keywords tab opens on Core. Switch the band filter to see the rest, and to rescue anything you disagree with by saving it to a collection.
Reading the report
Articles
The deliverable. Each article has a primary keyword, the supporting keywords it should also target, a working title and meta description, and a short explanation of why its primary keyword was judged core. Articles are built only from core keywords unless you switch on the peripheral toggle. Select articles and use "See these keywords" to jump to their rows in the Keywords tab.
Map
The topical map: hub pages and the spoke articles that should link to them. It is the site structure the article plan implies, useful for deciding categories and internal linking before anything is written.
Topics
Article clusters with their keyword and article counts, total search volume and average difficulty. Clicking a topic narrows both the article list and the keyword table to that cluster.
Keywords
The full table, one row per keyword, with volume, difficulty, CPC, intent and band. It is the same results table as the other research modes, so you can select rows and analyze them for Domain Scores, SERP weaknesses and traffic estimates, or save them to a collection.
Iterating on a report
- Redo articles regroups the saved keywords into a fresh article plan, with an optional re-rank, without harvesting again. It is charged at the depth's redo price (see the table above) because it re-checks live search results for every article.
- Rewrite titles rewrites only the article titles and meta descriptions, in the style you pick. The first rewrite per run is free; later ones cost the depth's rewrite price. Keywords, topics and articles stay exactly as they are.
- Duplicate plan starts a new plan from the same brief so you can change depth or constraints and run again.
Sharing with clients
Every report can be shared through a password-protected link. The client sees the same Articles, Map, Topics and Keywords tabs, including analyzed rows, without an account and without the ability to spend your credits.
Credits and refunds
Generating and revising a plan is free. Approving it charges the depth's price up front, and the run is settled when it finishes: whatever the run did not need is refunded, with at least 20% of the price retained on any run that produced results, to cover the planning and AI work. A run that fails or finds nothing is refunded in full. Redo articles is charged at the depth's redo price; the first titles rewrite per run is free and later ones cost the depth's rewrite price. Analyzing keywords from the report costs the same as analyzing them anywhere else. See How Credits Work for the full price list.
From the API or an AI agent
The whole flow is available over the REST API and the MCP server: create a plan from a brief, read and revise it, approve it, wait for the run and read every tab of the report. API keys carry their own spend limits and can be restricted to a maximum depth, so an agent can run research for you without being able to run up a bill.