“I need twenty backlinks, can you quote me?” The request almost always arrives in that order: the number first, the reason later.
My answer isn’t no. I buy links when the market justifies it, and I say so to my clients. Against soft competition, a few well-chosen spots at 30 or 50 euros do the job. Paying 300 euros for a big authority domain would be money thrown away. But a bought link is rented. It is never owned.
The other half of the plan is the half you don’t pay for. It is slower, it requires having something to offer, and it is the only half that stays with you.
1. What you buy, and what you never own
The French link market is organised, priced and fast. You pick a site from a catalogue, you pay, the article goes live. It is comfortable, which is exactly why the question arrives as a quote request.
Google’s own text has never been ambiguous. Its spam policies file buying or selling links for ranking purposes under link spam. The list is broad: money for links, sending someone a product in exchange for a write-up containing a link, paid advertorials whose links pass ranking credit. The same text leaves a door open. Buying and selling links for advertising purposes is a normal part of the economy of the web, and it is not a violation as long as those links carry a rel="nofollow" or rel="sponsored" attribute.
Read that door to the end. It says advertising is legitimate. It does not say a paid link passes anything. The documentation on qualifying outbound links states that sponsored is the value for advertisements and paid placements, and that it is preferred over nofollow. The day the publisher adds the attribute, you have bought advertising. An honest purchase, which can even bring qualified visits. It is no longer SEO.
And when a spam update lands on links, fixing things afterwards doesn’t help. Google’s documentation is blunt: once the effect of spammy links is removed, any ranking benefit they may have generated is lost, and it cannot be regained. The most recent one shipped on 24 March 2026 and rolled out in under twenty-four hours. You don’t see these coming. You build so you don’t depend on them.
There is a second front, less discussed in France. Since Google formalised site reputation abuse, a publisher hosting third-party content mainly because its own domain ranks well falls inside the rule. The text explicitly spares partner content whose purpose is to reach the publication’s readers, and targets content placed there to capture the host domain’s signals. So the line isn’t the presence of money, it’s intent. In a section sold as a package, the intent is written on the purchase order.
One last point, and this one is French law rather than SEO. Article 20 of the law on confidence in the digital economy provides that any advertising accessible through an online public communication service must be clearly identifiable as such. It must also make identifiable the person on whose behalf it is carried out. So the “partner content” label is not a publisher’s courtesy. On what exactly that text covers, a lawyer will answer better than I will.
2. The blocker isn’t the email, it’s what you have to offer
When a client tells me outreach isn’t working, we open the sent folder together. The problem is almost never the wording. It sits upstream: you are asking for a link without giving anyone anything to link to.
A journalist, a trade blogger, a professional federation don’t link to someone. They link to something, because that something makes their own piece better. Three things to own, then, before writing to anyone.
A number nobody else has
This is the upstream work, and it’s the part everyone skips. Your sales, your after-sales service, your rejected quotes produce numbers nobody else holds. Take a seller of professional kitchen equipment: four years of repair tickets tell him, by product family, after how many months a part gives out. No title in his trade press has that data. Several would run a piece on it.
One condition, and it disqualifies you if you miss it: publish the method alongside the number. The scope, the period, the number of observations. A number without its method isn’t quotable, it’s a claim, and a journalist putting their byline on a piece doesn’t repeat a claim.
Someone with a name
A publication quotes a person, not a brand. It needs a name, a role, a photo, and someone who picks up the phone on the day the piece closes. When the owner refuses to appear, that route is shut, and I say so before we start rather than after three months of emails. A proper author page on your site, with the person’s real track record, serves this route better than ten blog posts signed “the editorial team”.
A page that already exists
Live, indexed, and answering exactly the question asked. Not the homepage, not a sales page. Whoever reads you has to check in ten seconds that the link holds up for their own readers.
3. Three routes that don’t go through your wallet
| The route | What you need to own | Who answers | The trap |
|---|---|---|---|
| Proprietary data | An aggregated, anonymised, dated dataset, with its method | Trade press, sector blogs, professional federations | Publishing the number without the method: nobody can pick it up |
| Quoted expertise | A named, reachable person, with an author page | Journalists on deadline, podcasts, op-eds, schools | Delegating the answer to an agency: they want the practitioner |
| Your ecosystem | An existing relationship and a page that deserves the link | Suppliers, clients, associations, trade shows, local authorities | Asking for a link to the homepage: nobody knows where to put it |
The third route is the most neglected, and the cheapest. A manufacturer has resellers, a reseller has brands, a company has a federation, a trade show, a vocational school training its future staff. Those pages already exist, they already list names, and nobody has ever asked to be on them.
The most expensive mistake
Asking before the page exists. You email thirty sites, you land an agreement, and you discover you have nothing to point at. The link ends up on the homepage, which ranks for nothing in particular, and the opportunity is spent. You don’t go back to the same editor twice in one quarter. Page first, request second.
4. The request: five lines, one idea
The subject line names the thing, not the intention. The first line says why this person and not another: a piece they signed, an angle they follow. The second gives the number or the page. The third says what they can do with it. No attachment, no press kit, no “I am reaching out to you”.
One position I hold: I never offer to write the article. The moment you offer the text, the conversation changes nature and you are back on the advertorial side. You offer the material and leave the writing to whoever does it for a living. Anchors call for the same subtlety: a journalist writes in their own words, not in your target query, and an over-optimised anchor gets refused or stripped.
Worst case, nobody replies and you’ve lost one email. Best case, you enter the address book of someone who publishes every week and who will come back to you on their own next time.
5. Timing, and rhythm
There is always a right moment, and moving too early costs you the opportunity. A site launched last week, with no history and no in-depth page, has nothing to offer. It is sent straight to a refusal. It needs to mature.
Rhythm matters as much as volume. Two or three chosen requests a month beat a burst of thirty sent at once. A burst is spotted, it burns the contacts, and it produces a link profile whose curve looks like nothing natural.
6. What I watch to know whether it’s working
Not the number of links. Three numbers, and they don’t move at the same pace.
- Distinct referring domains. Ten links from the same site count as one domain. It is the only unit that means anything when you compare yourself to a competitor.
- The page that received the link. If it isn’t the one you wanted to push, the effort served something else. That check happens at publication, not at the quarterly review.
- The health of the site quoting you. An authority domain sending out far more links than it receives looks like a link farm. A site whose traffic has collapsed has already slipped through Google’s net, and its link will carry you nowhere.
Same rigour on the cleanup side: I never take a disavow file the way a tool exports it. You read it line by line. A genuine site regularly ends up flagged as toxic there by mistake.
In the crucible: telling a rented link from an earned one
A link budget buys a position for as long as you renew it. Your own data, an identifiable person and a page worth quoting produce links you stop paying for, and that search engines and assistants then reuse to name you as a source. Both are legitimate practices, and I practise both. Only one of them stays in the house when the payments stop.
The real work isn’t in the outbox. It is in what your company already knows and has never written down. The next stretch of the same road is how those links turn into citations inside AI answers, and how to pick the small editorial links when you do decide to buy. Without a content strategy behind it, there is nothing to point at.
Paying for links without knowing what they return?
Send me your domain name and the list of links you have already bought. I will hand back the real state of the profile: distinct referring domains, the ones already carrying a sponsored attribute, the sites whose traffic has collapsed since the purchase. And the step you have to climb against the three competitors ahead of you. Worst case you confirm your budget is well placed, best case you stop renewing a spend that carries nothing any more.