SMF·MANGOOLS
Can a prompt replace Mangools?
SEO & marketing — keyword research, backlinks and rank tracking
Exhibit tracking slip
Verdict
Every number Mangools shows you comes from somewhere you cannot get to: search volumes licensed from clickstream providers, a link index built by crawling the web at scale, SERPs pulled from a hundred countries every day. A personal build can work with the one dataset you genuinely own, which is your own Search Console data, and that is a useful tool. It is not this tool, and pretending otherwise would be the dishonest version of this entry.
Exhibit A — The prompt
Received on31.07.2026Build a personal keyword workspace on top of Google Search Console data, for a site you own.
Setup: register an OAuth client in Google Cloud, authenticate against the Search Console API, and let the user pick one or more verified properties. Store refresh tokens encrypted on disk.
Core loop: every night, pull the previous day's query-level rows (query, page, country, device, impressions, clicks, position) and append them to a local SQLite table. Search Console only keeps 16 months; this archive keeps everything from the day you start, which is the single most valuable thing this build gives you over the Search Console interface.
Build these views on top of that archive:
- Striking distance: queries sitting at average position 5 to 20 with meaningful impressions, sorted by impressions, since those are the ones a small edit can move.
- Cannibalisation: queries where more than one URL from the property has appeared in the results within the same week.
- Decay: queries whose 28-day impression trend is down more than a chosen percentage against the same window a year earlier.
- Question queries: rows matching who / what / why / how / can / does patterns, grouped by the page that ranks for them.
- Page clusters: for each URL, the full query set it ranks for, so a page's actual topic can be compared against its intended one.
Add a manual keyword list where the user can paste keywords they are considering, with a free-text note and a volume figure they type in themselves from whatever source they have. Label that column clearly as user-entered. Never invent, estimate or interpolate a search volume.
Out of scope: scraping search results, any third-party volume or difficulty estimate, backlink data, and competitor analysis. This tool reports on one property's own measured performance and says so.Opening prefills the prompt — press enter to run it.
Exhibit B — What you lose
- B.1 search volume for keywords you do not already rank for
- B.2 keyword difficulty scores
- B.3 the backlink index behind LinkMiner
- B.4 SERP data for countries and devices outside your own traffic
Prior art
Exhibit C — Why people still pay: proprietary web index and data acquisition
Because Search Console only tells you about pages you already rank for. The entire point of keyword research is finding the ones you do not, and that requires a dataset somebody spent millions collecting.
Questions
Can I import my Mangools keyword lists?
You can paste them into the manual list, and the KWFinder CSV export maps cleanly onto keyword plus volume plus note. The volumes come across as user-entered values with the date you imported them, because they will go stale and nothing here refreshes them.
Why not just use Search Console directly?
Two reasons. It discards data after 16 months, and this build keeps it forever. And its interface caps what you can compare, so questions like "which queries did this page lose between two redesigns" are painful there and a single query here.
What does it cost to run?
Nothing. The Search Console API is free with generous quotas for a personal site, the data is small enough to sit in SQLite for years, and a nightly pull runs fine on any always-on machine or a small VPS.
What is the honest gap against the paid tool?
Discovery. This tells you everything about the keywords you already touch and nothing at all about the ones you do not, which is the entire question keyword research exists to answer. Nothing in a personal build closes that; it needs a purchased dataset.
Related tools
Receipt