Proposals are written. Case studies are curated. Live placements are neither — they are public artefacts that exist independently of how anyone describes them. Which makes them the only genuinely reliable evidence available to you before money changes hands.
So ask. The request is simple and entirely reasonable: twenty live URLs from placements made in the last ninety days, client names redacted if necessary. An agency that will not produce these has told you something useful. An agency that produces them enthusiastically usually knows they will survive inspection.
Before you start
You need a backlink tool with traffic estimates — Ahrefs, Semrush or similar — and about ninety minutes. If you do not have a subscription, most offer a trial that covers this comfortably, and the exercise is worth the cost of one month regardless.
Set up a sheet with one row per URL and a column for each check below. Resist the urge to form an impression from the first three; the pattern across twenty is the thing you are looking for.
Check 1 — does the page still exist, and is the link still there?
Open each URL. Find the link. Confirm it points where it should and is not tagged rel="nofollow" unless that was disclosed.
Some attrition is normal — pages get pruned, sites get sold, editors revise. Across twenty placements from the last ninety days, though, you would expect nearly all of them intact. If three or four are already gone, that is not attrition, that is the inventory.
Check 2 — does the referring domain have readers?
Run each host domain through your traffic tool. You are looking at organic sessions, not domain rating.
| Monthly organic traffic | Read as |
|---|---|
| Under 500 | Effectively unread; the link is decorative |
| 500–1,000 | Marginal; fine occasionally, alarming if it is the pattern |
| 1,000–10,000 | Normal working range for niche trade publications |
| 10,000+ | Genuinely valuable if the topic is relevant |
Watch for the specific combination of high domain rating and negligible traffic. That pattern is the signature of a site built to sell links, and it appears in bulk in the cheaper tiers of this market.
Check 3 — did the site cover this topic before the agency arrived?
Look at the publication's archive. Was it writing about your category two years ago, or did it pivot into "business technology" recently and start publishing a wide range of unrelated contributed articles?
A publication that covers everything covers nothing, and a link from it carries the topical weight of a directory listing.
Check 4 — who wrote it, and do they exist?
Click the byline. Is there an author page? Do they have other articles, a professional profile, a footprint outside this one site? An anonymous "admin" byline on a WordPress install with a stock avatar is a footprint you inherit.
This check takes about thirty seconds per URL and catches more low-quality inventory than any other single test.
Check 5 — where does the link actually sit?
Read the paragraph around it. Apply the deletion test: if you removed the sentence containing the link, would the article be worse?
- Good: mid-article, inside an argument, where the link supports a claim being made.
- Acceptable: in a genuinely relevant list of resources.
- Weak: author bio, footer, sidebar, or a "sponsors" block.
- Bad: a paragraph that exists solely to accommodate the link and reads like it.
Check 6 — is the content any good?
Read two of the articles properly. Not skim — read. You are assessing whether a competent editor would have published this without payment.
This matters commercially beyond link quality: these articles carry your client's name, and soon yours. If the writing is thin, generic and clearly produced at volume, that is what will appear under your brand.
Check 7 — do the twenty domains look like each other?
The pattern check, and the one people skip. Look across all twenty for shared characteristics: the same WordPress theme, similar site architecture, near-identical "write for us" pages, comparable outbound link profiles, hosting on the same IP ranges.
Individually, none of these proves anything. Together, they describe a network — and a network is exactly the thing you are paying an agency to avoid.
Scoring it
Out of twenty placements, a healthy result looks roughly like this: 18+ still live, 16+ with over 1,000 monthly organic sessions, 18+ with a real named author, 15+ contextual in-body links, and no visible footprint clustering.
Below those numbers, ask about the gap directly. The explanation may be reasonable — a difficult category, a client with an unusual budget — but you want to hear it before you sign rather than discover it in month five.
The two questions to ask afterwards
Whatever the audit shows, go back with two things.
"Three of these have under 500 monthly visitors. Walk me through why they were approved." A good agency will either defend the choice with a reason you had not considered, or acknowledge it. Both are fine. Deflection is not.
"Can I see the domains you rejected in the same period?" The twenty URLs show what got through. The rejection log shows where the bar sits — and it is a far more revealing document.
We have been on the receiving end of this audit perhaps a dozen times. It has never once been unwelcome, and the clients who ran it have been our longest engagements. People who check things tend to stay.
If they refuse
Some agencies decline, citing client confidentiality. Confidentiality is genuine and NDAs are standard practice — but published articles are public documents that anyone can read. Redacting the client name protects the relationship; withholding the URL protects something else.
If the answer is no, ask instead for twenty URLs from any client who has consented. An agency working with more than a handful of companies will have at least one.