How to get quality backlinks without buying them

Nobody links out of generosity. An editor links when they need to back up a sentence they just wrote. That explains why the most-taught tactic, the cold email to a stranger, is the one that ages worst: it asks for a full evaluation and offers nothing in return. What works inverts the order. First there is the material someone else finds it convenient to cite.

How do you get quality backlinks without paying for them?

Through slow work with no shortcut: giving others reasons to link. A new site gets more from ordering its internal links; a mid-authority site can start earning external ones; a consolidated site is the only one where paying for links is worth the price. Buying links that pass ranking signals is named in Google's spam policies.

Quick answer

• A link is not asked for, it is provoked. A cold email asks a favor; a figure of your own saves the writer work.
• The unlinked mention is the cheapest route. Someone already decided you were relevant and an edit is missing.
• The currency is hours, not money. The real constraint is how many afternoons fit in a quarter.
• Domain authority changes the pace, not the method. A new site runs the same routes with more silences.

What kind of problem this is

Almost everyone treats this as a sales job. You prospect, you write, you follow up, and the metric watched is the number of contacts sent.

It behaves more like a production problem. The link is not the product: it happens when someone writes about your topic and needs to back a claim with something they do not have. From that comes the variable that orders the work when there is no budget, which is hours. Two equally legitimate tactics are separated by an order of magnitude in hours per link.

Before any of it there is a one-afternoon check. In a 312-page recipe site I audited, the two pages with the most impressions received no internal link at all, and 79 % of the links that existed sat in the footer. An external link that lands there stays where it fell. The full order of operations is in internal linking.

The citable asset: the only route that compounds

A citable asset is an object with a number inside it that is not published anywhere else. It is not a well-written opinion. It is a measurement: a count, a time series, a table compared by hand. An example: an autocomplete harvest across Colombia, Spain and Mexico over 11,220 queries returned 14,715 unique suggestions, of which 8,875 survived cleaning, the kind of count that comes from no tool and that free keyword research explains how to produce.

The same material feeds two channels that are converging. The Princeton GEO study (arXiv 2311.09735, KDD 2024) tested nine methods over roughly 10,000 queries: adding statistics produced gains of 30 to 40 % in visibility inside generative-engine citations, and citing sources lifted the visibility of pages sitting around position 5 by 115 %. That measures machine citations, not human links, but the object that provokes both is the same, which is the argument in how to get cited by ChatGPT and the whole point of generative engine optimization.

The hours depend on how much of the road is done. With the data already inside Search Console, extracting, verifying and writing run 8 to 12 hours. If it has to be collected, 20 to 40. If the instrument has to be built first, it passes 60 and may yield nothing publishable. Form decides the rest: a number buried in the fourteenth paragraph is not extracted; in a table with its date beside it, it is. That depends on how an article is written to rank. The route goes dark in two cases: when there is no figure another cannot pull with the same tool, and when the figure belongs to a client with no written permission.

The unlinked mention: half the work is already done

Someone named the brand in a text and did not add the link. The editor already did the expensive part: deciding the name was worth including and writing the sentence around it. What is left costs twenty seconds. You do not ask for an evaluation, you point out an omission.

Candidates come from simple queries. The name in quotes with the -site: operator drops your own pages. The author's name finds texts the domain name does not. A reverse image search locates whoever reused an image without attribution. The message that gets answered is short: the exact URL, the exact sentence, and where the link would go. Half an hour per case. What to expect depends on how much of your name is out there: a freshly published site usually finds zero candidates, and there the tactic does not exist.

Someone else's broken link: filling a gap that already bothers them

A resource page links to something that stopped existing. The broken link is a visible defect on that page, not a pending favor, and that changes the exchange. Without paid tools, the path is to crawl the resource lists of trade bodies and universities, which rot fastest, with Screaming Frog, whose free version reaches 500 URLs.

The check that saves the afternoon is what "broken" means. A 404 is a real gap. A redirect to the home page is a link the browser resolves and the editor does not perceive as broken. A domain that lapsed and was rebought by someone else is worse: the destination exists and says something different. Telling those three states apart is technical SEO applied to someone else's page. The most frequent outcome is not the link but the editor deleting the line, because it costs less than evaluating a replacement.

Being where you are already named

There is a short list of organizations with a verifiable relationship to you that publish a page where your name fits without forcing anything: suppliers with a client page, clients with a supplier page, trade bodies and chambers of commerce, the university where you gave a talk, the certified-professionals directory of a tool you use.

What separates this from mass submission is the admission criterion. A directory that verifies activity, charges a fee and reviews by hand admits some and rejects others, so being inside says something. A form that publishes anything says nothing: a signal anyone can emit distinguishes no one. Each listing runs 1 to 4 hours plus any fee. The limit is where the fee stops buying a membership and starts buying the link.

What gets done a lot and moves nothing

Four tactics take most of the time and none survives a look at its mechanics. Mass link exchanges make the link a function of the deal, not the content, and the pattern is written into the data, two sites pointing at each other with anchors chosen by the interested party. Blog comments produce links the platform itself marks as nofollow or user-generated, since the common systems add rel="nofollow" or rel="ugc" by default. Generic directories fail for the reason above: no admission criterion, no evaluation, no signal. Distributed press releases publish the same text across dozens of sites, the opposite of digital PR built on a real finding, and optimized links inside distributed releases are among the link-scheme examples in Google's documentation.

The argument against them is not fear of a penalty. It is the month: four afternoons on an exchange are four afternoons not spent on the citable asset.

When NOT to do this

When the asset is not published yet. A contact is spent once, and writing before you have something to show spends it for nothing.

When the figure behind the piece belongs to a client with no written permission. The publishable version exists, unnamed and with rounded figures, but that is decided before writing.

When a core update is rolling out. The March 2026 update moved 79.5 % of the top 3 according to SE Ranking, so any measurement of a campaign in those weeks is unreadable.

When the only route available is emailing strangers. That is not a campaign with a poor response rate: it is the sign that there is nothing to cite yet.

Mistakes that repeat

• Asking for the link before publishing the asset. The contact is spent once, and spent with nothing to show.
• Writing to the generic inbox. An info@ address is read by someone who cannot edit the page.
• Sending the request without saying where exactly. An editor who has to hunt for the paragraph closes the email.
• Counting emails sent instead of hours invested. Fifty messages nobody read are an afternoon, not a campaign.
• Rewriting the page after earning the link. You were linked for a sentence; if it goes, the link backs nothing.
• Searching only for the domain name. The author's name and your own images surface candidates the URL misses.

Data and transparency

The hour costs are trade estimates, not published measurements, which is why they are given as ranges. No outreach response rate is cited: the ones that circulate are not auditable. The 30-to-40 % gain in citations from adding statistics and the 115 % for pages around position 5 from citing sources are from the Princeton GEO study (arXiv 2311.09735, KDD 2024). The 11,220 queries, 14,715 suggestions and 8,875 useful come from an own autocomplete harvest across Colombia, Spain and Mexico. The 312 pages and the 79 % of internal links in the footer come from an own audit of a client site that is not named, with rounded figures. The 79.5 % of the top 3 displaced in the March 2026 core update is from SE Ranking. That the common comment platforms add rel="nofollow" or rel="ugc", and that optimized links in distributed press releases are a link-scheme example, come from Google's public documentation. Google does not publish the weight of links as a signal. Verified as of September 2026.

Primary sources, opened 16 September 2026: the Princeton GEO study; the SE Ranking survey.

What this changes

Getting links is framed as a channel to work. Seen from production, what decides the quarter is something else: what you published that someone could need to back a sentence of their own. When there is nothing, no volume of emails corrects it.

The two channels that hand out visibility now ask for the same thing. An editor links the figure because it saves them proving it, and a generative engine cites it for the identical reason. Writing to earn the link is already writing for the citation. An email can be automated, and that is why it is worth what it is worth. A measurement of your own cannot, and that is why it gets linked.

Frequently asked questions