SMF·KEYSEARCH
Can a prompt replace Keysearch?
SEO & marketing — keyword research, backlinks and rank tracking
Exhibit tracking slip
Verdict
Keysearch is bought for one number: how hard is this keyword. That score blends link metrics, domain authority and on-page signals, and the link half of it comes from a crawl index nobody can rebuild. What you can compute honestly is the on-page half — from a SERP you paid an API for, how many results actually target the phrase, how long they are, how concentrated the domains. That is a partial difficulty score, it is transparent, and it is genuinely useful as long as it never pretends to be the whole picture.
Exhibit A — The prompt
Received on31.07.2026Build a keyword workspace whose difficulty score is computed from observable signals and shown as an open formula.
Lists: keywords organised into named lists with tags and a free-text note. Import from CSV. A keyword carries a target country, language and device.
SERP collection: fetch the top 10 organic results per keyword through a commercial SERP API behind a swappable provider module, never by scraping. Enforce a per-day query budget and show spend on the dashboard. Cache by keyword plus locale plus day; never re-fetch the same keyword twice in a day.
Page signals: for each result, fetch the page itself, politely and honouring robots.txt, and extract the title, H1, word count, whether the exact phrase appears in the title and in the H1, and whether the URL contains a slugified form of the phrase.
The score, which is the point of this build. Compute a 0-100 difficulty from these components, each weighted and each shown:
- Title targeting: what share of the ten results have the phrase or a close variant in the title. High share means the keyword is being competed for deliberately.
- Depth: the median word count of the ranking pages, normalised against a configurable expectation.
- Domain concentration: how many distinct domains hold the ten positions, and whether any single domain holds several.
- Result type: how many positions are taken by user-generated platforms, aggregators or news, which usually indicates a softer target.
- Home-page share: how many results are a domain's home page rather than a specific article, which usually indicates the keyword is broader than it looks.
Display the score as a bar plus a breakdown table with every component's raw value, its weight and its contribution. Let the user change the weights and see every keyword rescore live. A score you cannot audit is a score you cannot argue with.
State explicitly, in the interface and not only the README, that this score contains no link or authority data, that it therefore systematically underestimates difficulty on established topics, and that it is a shortlist tool rather than a verdict.
Also build: bulk scoring of a whole list with progress and budget enforcement, sort and filter by any component, export to CSV including the breakdown, and a rescore that reuses cached SERPs so weight experiments cost nothing.
Out of scope: search volume from any source, estimated or otherwise; backlink data; rank tracking over time; and any keyword idea generation.Opening prefills the prompt — press enter to run it.
Exhibit B — What you lose
- B.1 search volume, which no free source provides for keywords you do not rank for
- B.2 the link and authority half of any real difficulty score
- B.3 the keyword database you browse for ideas
- B.4 YouTube and competitor research bundled in the same subscription
Exhibit C — Why people still pay: proprietary web index and data acquisition
Because twenty-four dollars buys a volume figure and a difficulty number for any keyword you can think of, and reproducing even half of that costs more in API calls than the subscription does.
Questions
Can I import my Keysearch lists?
Yes, as CSV, and keywords, tags and your own notes come across. Keysearch's difficulty numbers do not import, because this build computes a different score from different inputs and showing two incomparable numbers side by side would be misleading.
How close is this score to a real difficulty metric?
Directionally useful, systematically optimistic. Without link data it cannot see that a page ranks because a thousand sites link to it, so it will call an established topic easier than it is. It is reliable for separating obviously soft targets from obviously hard ones and unreliable in the middle.
What does it cost to run?
The SERP API dominates: one to three dollars per thousand queries, so scoring 300 keywords is under a dollar, plus the page fetches which are free. Against a monthly subscription this wins clearly for occasional research and loses for daily use.
What is the one thing that does not survive the rebuild?
Volume. Difficulty without volume tells you a keyword is winnable but not whether winning is worth anything, and there is no legitimate free source for search volume on a keyword you do not already rank for.
Related tools
Receipt