backlink-guard

backlink-guard

Check a link source before you submit to it, the link after it goes live, and your own backlinks for link spam.

⭐ 37 of 40 Ahrefs-marked spam sites kept from "submit", on 100 sites it had never seen · on PyPI, the official MCP Registry and skills.sh

Quick start · How well it works · Where to be · Competitors · AI answers · Keys · Agents and MCP · 中文说明

Find the pages your searchers already land on, the sites that link to your competitors and not to you, the sites Google's AI Overviews cite for your keywords, and what AI assistants answer when asked for what you make; check a link source before you submit your site to it, check the link after it goes live, look at your own backlink profile for the patterns Google calls link spam, and find the links that point at pages of yours that are gone or redirect badly.

Most backlink tools help you send more. This one tells you where not to send, and when to slow down.

Quick start

uv's uvx runs it straight from PyPI, with nothing to install; uvx backlink-guard@latest … picks up a new release. To keep a backlink-guard command installed: pip install backlink-guard (or uv tool install backlink-guard). Python 3.10 or newer, no dependencies. With no key it reads the sites' own pages, for free (--free asks for just that). The keys (GoAnyAPI, and Glasser or AIsa) add rating, visits, Google's index and spam scores: about $0.17 for one site and $0.46 for thirty, so check the sites you are considering in one run.

Checked with both keys on 2026-10-13 and 14: Action is what to do with the site: skip, look first, free only, low value or submit. It reads three answers that are also shown apart. Link policy: what a link there would mean under Google's link spam policies, one finding being enough; mcpservers.org sells followed links, so list there for free and never pay. Value: what a listing can bring, which is the site's reach, or none when nobody can reach it (dead, parked, dropped by Google, a shortener, a network page). The verdict weighs every finding: one avoid finding makes avoid, and caution takes two caution findings. Risk is 3 for each avoid finding and 1 for each caution. A site its own pages put at avoid, like navi.tools, gets no paid check.

Every finding comes with the text or the numbers it was read from: turbo0.com — ⚠️ A high rating with almost no visitors: the rating comes from links, not readers. *rating 81, about 2,218 visits a month.* ⚠️ It asks for a link or badge back as a condition. *"Free $0 … Backlink to our site is required."* (Google: Excessive link exchanges, or requiring a link as part of an arrangement)