🔗 How to Find Backlinks in 2026

Start here: your own site, free and authoritative

If the site is yours, skip everything else on this page first. The two engines that removed public backlink operators will still hand you their own link data — you just have to prove you own the domain. This is better data than any third-party crawler can give you, because it's the index itself talking.

The graveyard: why link: is gone

Every engine that offered a backlink operator has removed it, and they removed it for the same two reasons: serving link-graph queries is expensive, and the answer is competitive intelligence that the engine has no incentive to give away.

The lesson isn't "search engines got worse." It's that a public backlink operator is fundamentally at odds with an engine's interests — it leaks ranking signals and it invites scraping. Engines would rather leave that market to Ahrefs and Majestic than hand it out for free. Nothing about that calculus is going to change, so stop waiting for link: to come back.

What actually survives (free, any domain)

Marginalia is an independent Swedish-built engine with its own crawler — it doesn't reuse Google's or Bing's index — and it's open source. Crucially, it still exposes a per-domain backlinks view:

https://marginalia-search.com/site/yoursite.com?view=links

That shows links to the domain from other sites in Marginalia's index. You can wire it up as a browser keyword shortcut with %s in place of the domain and have a one-keystroke backlink check.

Mojeek (another independent crawler) has no link: operator, but it does document inanchor:, which finds pages linked with specific anchor text. That's a partial proxy: you can't ask "who links to me," but you can ask "who uses my brand name as anchor text," which catches a useful slice of the same thing. Someone did ask Mojeek for a referrer: operator — it was a feature request, not a feature. Operator docs: mojeek.com/support/search-operators.html.

Not a backlink tool, but it finds the same links from the other end. Search your exact brand or domain as a phrase, excluding your own site, and you surface pages that mention you — a large share of which link to you:

"yoursite.com" -site:yoursite.com
"Your Brand Name" -site:yoursite.com

Add after:2026-01-01 to catch new ones. Set it as a Google Alert and it becomes passive link monitoring for free. This works on Google, Bing, Mojeek, and Marginalia alike, because it uses operators nobody has any reason to remove. See our Google dorking reference for the full operator set.

Commercial services: what you're actually buying

You're paying for a crawler and an index, not a search feature. The four that matter run their own crawls at a scale nobody else replicates. Prices and index figures below are as published Q1–Q2 2026.

• Link building is your job → Ahrefs. Freshest index, cleanest UX — median ~3.2 hrs to surface a new link. If you only buy one tool and links are the work, this is it.
• You span SEO + paid + content → Semrush. Widest feature surface; backlink data connects to rankings and keyword modules in one place.
• Forensic link work — disavows, penalty audits, PBN mapping → Majestic. Trust Flow / Citation Flow are the sharpest quality metrics available, and the Historic Index reaches back to 2006. In one documented penalty audit it surfaced 340 links the other tools missed entirely. At $49.99 it's also the cheapest way into serious link data.
• Client reporting to non-technical people → Moz. DA is the metric clients have heard of. That's genuinely the main argument for it.

The DIY route: Common Crawl's webgraph

This is the only place you can genuinely "dork all backlinks," because you're not asking an engine to be generous — you're querying the raw link graph yourself. Common Crawl publishes the full hyperlink graph of its crawls as three plain files per release:

Finding your backlinks is then four passes: resolve your domain to a vertex id, sweep the edge list for every from_id pointing at it, turn those ids back into names, and optionally attach ranks. It's public data — no ToS problem, no API key, no rate limit. The cost is bandwidth: you filter locally, so a full run pulls roughly 17 GB.

You could do this with zcat | grep and enough patience. The tool below does it with a progress bar, resolves ids to names, attaches authority scores, and exports CSV.

The tool: Max Intel Linkgraph

A free, open, offline desktop app that pulls every referring domain for any site out of the Common Crawl webgraph. It's the closest thing to a working link: operator that still exists in public. Nothing is sent to us — it talks directly to Common Crawl and filters on your machine.

1. Install Python if you haven't got it — python.org/downloads. On Windows, tick "Add Python to PATH" in the installer or nothing below will work.
2. Install the two bits it needs. Open Terminal (Mac) or Command Prompt (Windows) and paste:

   pip install customtkinter requests

3. Run it:

   python maxintel-linkgraph.py

   A window opens. No install, no account.
4. Type a domain (example.com) and press Find backlinks. It'll warn you about the download size and ask you to confirm — that prompt is the last exit.
5. Wait. Step 2 ("sweeping the edge list") is the long one — that's the 14.6 GB file. The progress bar is real; leave it alone.
6. Read the results. Every referring domain, sorted by authority. Double-click a row to open that site. Export CSV for the full list.

If it says your domain isn't in the graph: that means Common Crawl didn't reach it in that crawl, or nothing links to it yet. Try an older release from the dropdown before concluding anything. Small and new sites are frequently missing — that's a limit of the sample, not a verdict on your site.

Known quirk (July 2026): Common Crawl's release-list endpoint (index.commoncrawl.org/graphinfo.json) has been returning 503. The app handles it — it falls back to known-good releases and tells you it did, and lookups work normally. If the dropdown looks short, that's why.

Frequently asked questions

Related