Undoing Four Years of Cheap Links

They arrived four weeks after a manual action, having lost about 70% of organic traffic overnight. The previous agency had spent four years buying links — 400-odd domains, most of them from the same handful of private networks, all pointing at the same six pages with the same four anchor phrases.

The obvious move was to disavow everything and start again. That was also the wrong move: buried in the profile were roughly 90 legitimate links earned before the agency arrived, including two from publications that would never link again if the relationship soured. Blanket-disavowing would have cost them the only real authority they had.

The harder constraint was commercial. They had a funding round closing in two quarters and the traffic line was in the deck. Recovery had to be real, and it had to be explicable to people who were not going to read a backlink audit.

Every referring domain got looked at individually. No toxicity score, no automated threshold — those tools flag legitimate sites constantly and miss the well-built networks entirely. It took nine days and produced three buckets: 412 to disavow, 90 to keep, and 58 genuinely ambiguous ones that got a second pass with the client in the room.

For 30 of the worst offenders we requested removal before disavowing. Eleven complied. That is a poor hit rate and roughly what anyone should expect, but documented removal attempts materially strengthen a reconsideration request, and the request was the actual deliverable.

The reconsideration request was written as a narrative rather than a form submission: what happened, when, who did it, what has been removed, what has changed structurally so it does not recur. It was accepted on the second submission — the first came back asking for detail on a set of domains we had classified as ambiguous and left live. We disavowed them and resubmitted.

Only after the action was lifted did we build anything new. Twenty-two editorial placements over the following ten weeks, deliberately slow and deliberately boring, pointed at the pages that had lost the most. A profile under review is the worst possible moment to start acquiring links quickly.

The manual action was lifted in month three. Traffic reached 98% of pre-penalty levels by month five, and passed it in month seven — after the engagement had formally ended, on the back of the new placements maturing.

412 domains were disavowed and 90 kept. The kept ones matter more than the removed ones: two of them still rank the pages they point at, and neither would have survived a blanket file.

The honest caveat is that recovery of this kind is not guaranteed and the timeline is not controllable. Five months was a good outcome, not a standard one. Anyone quoting a fixed recovery window for a manual action is quoting a number they cannot influence.