Outreach link building services compared with placement brokers, and how to tell them apart
Both sell links and they are not the same purchase. One asks publishers; the other has inventory. Four questions separate them before you sign, and Google publishes a sentence about the second.
Published by AI Knows Us · Updated 21 October 2026
Two very different businesses answer an enquiry about outreach link building. The first contacts publishers it has no relationship with, pitches something, and reports what was accepted and what was refused. The second already has a list of sites that will publish anything for a fee, and the outreach is an internal purchase order. Both will describe themselves as outreach. One of them has inventory.
We sell neither. We are not an agency, place no links and have no retainer, so this is a buyer's guide rather than a shortlist. The criterion is the same one throughout: what can the supplier show you in writing before you commit?
The four questions that separate them
| Question | Real outreach answers | A broker answers |
|---|---|---|
| Can I see the last month of sent emails? | Yes, redacted if needed. The pitch is the product and they are usually proud of it. | Vaguely. There may be no outreach to show, because the relationship predates your brief. |
| What is your refusal rate? | A number, and usually a high one. Real pitching is refused most of the time and a supplier who says so is describing the job accurately. | Near zero, which is the tell. Inventory does not refuse. |
| Can you name three publications you have never placed on? | Easily, often with the reason. | Awkwardly, because the universe of targets is the list. |
| Who pays the publisher? | Often nobody. The pitch was accepted on merit, or a fee is disclosed line by line. | They do, and it is inside your retainer rather than itemised. |
What Google publishes about the broker model
Google's spam policies page, read on 21 October 2026, defines link spam as the practice of creating links to or from a site primarily for the purpose of manipulating search rankings, and lists among its examples buying or selling links for ranking purposes. It also names excessive link exchanges or partner pages existing only for cross linking, and requiring a link as part of a terms of service, contract or similar arrangement.
That is published by the engine whose rankings the purchase is for, so it is the document to read before the agency's deck rather than after. It does not make a paid placement impossible to buy; it makes the risk a thing you have chosen with your eyes open rather than a thing you were told was white hat.
What a real outreach engagement produces each month
Five line items, and the third is the one that proves the work happened.
The URLs won, with the page of yours each one points at and whether the link is followed. The prospects contacted, with the publication named. The refusals, with the reason where one was given, because a month of outreach with no refusals is a month of something else. The assets used, meaning the data, story or tool that gave the pitch a reason to exist. And the anchors, because anchor text drifting towards the same commercial phrase on every placement is the pattern that makes a profile look bought.
Ask for that shape in the first month rather than the sixth. A supplier who can produce it has a process; one who produces a count and a chart has a spreadsheet.
What makes outreach work, and it is not the sending
The variable that moves reply rates is not the sequence, the subject line or the volume. It is whether the thing being offered is worth publishing. An email saying your readers would find our guide useful competes with forty identical emails that week. An email offering a statistic nobody else holds, or a correction to something the publication got wrong, or a tool their readers would actually use, is a different email.
This is why the asset question belongs in the contract. If the supplier is expected to earn links with nothing from you, the only remaining route is the one with inventory.
Where a supplier beats us, and where we are not in the comparison
We are not in this comparison at all. We do not place links, run no outreach, hold no publisher relationships and sell no retainer. A supplier who can show you sent emails and a refusal rate is the right purchase here and nothing on this page argues otherwise.
Our rendered pricing page also carries no SSO, no white label, no SOC 2 and no ISO line, and covers one business per plan outside the agency tier, which covers ten. For an agency buying on our behalf, Rankscale publishes the removal of its branding from shared dashboards and we publish no white label at all, which is a straightforward win for them.
The outreach result nobody is reporting yet
There is a second outcome of good outreach that no monthly report currently contains. When a publication writes about your category and names you, that page becomes a source an assistant can read. Months later, when a buyer asks ChatGPT or Gemini who to use, the answer is assembled from sources like it, and the businesses named in those sources are the businesses that get named in the answer.
That is the part worth measuring, and it is measured by asking rather than by counting links. AI Knows Us asks the buyer questions you choose across ChatGPT, Gemini, Perplexity, Claude and Google AI Overviews, reports which businesses the answer named, and lists the sources behind each answer, which tells you which placements are doing work and which are decorative. The published arithmetic, from the rendered pricing page: one credit is one buyer question asked on one AI app once, most checks use one credit, a Claude check uses 3, and a failed check refunds its credits. llms.txt publishes the rule that keeps it honest: the buyer questions never name the business being measured. The US floor is $149 a month for SEO Start, which llms.txt states watches 0 buyer questions on 0 engines, so Growth at $299 is the cheapest plan reading all five surfaces.
Ahrefs runs the same check and beats us on two published counts: its Brand Radar page states each prompt is checked daily on every selected platform, against our weekly on priority questions, and it reads AI Mode and Copilot, which we do not. Rankscale lists nine engines against our five. The one criterion where our published figure is the most exact is the weight of the expensive engine: Ahrefs publishes 8 checks per update for Claude, Rankscale publishes a flat typical 0.25 credits per engine per prompt with no per engine weight, and we publish the exact credit weight.
If you want the two products set side by side on what each one publishes rather than on what each one claims, that comparison is written out in our piece on AI Knows Us against Ahrefs Brand Radar. We measured durable inbound links ourselves on 21 October 2026, reading 2,169 stored bodies (1,797 drafts plus 372 fetched) against all 2,082 pages tracked on this site: 932 of them, 44.8% of 2,082, carry no inbound link written into another page's body. No other page on this site links to that one.
Questions buyers ask about outreach link building services
How do I tell outreach from a placement broker?
Ask for the refusal rate. Real pitching is refused most of the time, so a near zero rate means the targets were already arranged. Then ask to see a month of sent emails.
Should I pay per link or on a retainer?
Per link rewards volume and pushes a supplier towards inventory. A retainer with a named deliverable list rewards process. If you buy per link, specify the publications rather than the count.
Is buying a placement against Google's policy?
Google's spam policies page names buying or selling links for ranking purposes as an example of link spam. Read it directly before signing, so the risk is one you chose.
What does good outreach need from me?
Something worth publishing. Data from your own operation, a free tool, a correction, a story. Without one of those the supplier has only its inventory to fall back on.
What should outreach reporting include that it usually does not?
Whether the coverage changed who the assistants name when a buyer asks who to use. That is the outcome a cited article actually produces now, and counting links does not show it.