Nine months of nothing but branded anchors — posting what actually happened

Anyway this is same in every discussion here. Somebody test one thing and change four.

reading this instead of doing my internal links

Pulled what I can from my own side of the desk, which is not a study but is a lot of transactions.

Of placements I sold in the last two years, roughly 70% now specify branded or bare URL anchors. Two years before that it was under 40%. So the behaviour change is real and it happened fast.

What I can't do is connect that to outcomes, because clients don't tell me what happened and the ones who come back and buy more are a self-selecting group.

One thing I will say: the buyers who still insist on exact match are overwhelmingly the ones running programmatic sets and affiliate roundups, and they are also the ones who churn fastest. Whether that's the anchors or the model, I couldn't tell you. Depends what you're running, and what they're running is fragile regardless of what's in the link.

70% against 40% is the first hard number in this thread that isn't one man's site. Do you have it by month?

Q1 update, and it complicates my own post.

Quarter	Links bought	Content added	Median position, start	End
Q2 25	14	12 pages	18.4	16.1
Q3 25	17	9 pages	16.1	14.9
Q4 25	13	19 pages	14.9	11.2
Q1 26	0	0 pages	11.2	10.6

I deliberately bought nothing and published nothing this quarter, partly because of this thread. The median still improved, by less, through a core update in February.

Two readings. Either the Q4 links were still maturing and I'm seeing lag, which is the flattering version. Or the whole nine-month curve is substantially the site aging and the vertical moving, and the 44 links were a smaller part of it than I assumed when I opened this thread.

I genuinely don't know which. I'm inclined towards the second, which means my opening post is worth less than the 29 thanks on it suggest.

Had a quarter like that in 2022 - stopped everything on a site for four months because I was moving house and honestly forgot about it, and it climbed the whole time.

Drove me mad for about a year afterwards. I'd spent two years attributing every movement to whatever I'd done that month, and then it moved most when I did nothing at all.

What I took from it - and this is not scientific - is that most of what I do is probably noise sitting on top of a slow trend I can't see. Doesn't stop me doing it. Does stop me claiming credit.

Have you got a site you've genuinely left alone for a year to compare against?

so nobody knows if anchors do anything

reporting back on my exact match test because i said i would and nobody else has posted anything

pointed 6 links at one page, all exact match on the phrase, over february. page went from 24 to 19 and then back to 23 after the march update. two other pages i didn't touch went from 21 to 17 and 30 to 26 in the same window

so the pages i did nothing to beat the page i spent 700 quid on. i don't know what to do with that information

You do know what to do with it. You just don't want to.

That's the most useful post in this thread, incidentally, and it's from the person everyone here corrects. n=1, no controls, and it still contains an actual comparison group, which is more than my post or MetricsOverMagic's opening had.

Seven hundred quid for a negative result you published. That's a better contribution than nine months of a curve with no counterfactual.

Seven hundred pounds of evidence and it came from the man everyone corrects.

Two link-selling posts removed. Byte_broker's signature is fine, an unsolicited price list is not.

Ran PipelineRat's internal design after the March update settled, because this thread deserved at least one test somebody actually did.

84 pages on a client site, split into two matched groups of 42 by current impression volume. Group A had internal anchors rewritten by hand — same source pages, same link structure, only the words changed. Group B untouched. Eight weeks, no content changes to either group, no external links to either.

Group A impressions up 14% against group B's 6% over the window. Positions barely moved for either.

That is a smaller effect than I would have predicted in January and it is on impressions rather than positions, which usually means the pages became eligible for more query variants rather than ranking better for the ones they had.

Caveats, and they're heavy: one site, one vertical, and I did the rewriting myself, so old_growth_seo's attention objection applies in full. A person who knows the site read 42 pages carefully. That's the intervention. The anchors are just what I typed while doing it.

So after five months I can report an 8-point difference that I can't cleanly attribute, which is roughly where this thread has been since post 3.

Closing my end at fifteen months. Q2 I resumed buying — 11 links, branded only again — and added no content.

Quarter	Links	Content	Start	End
Q1 26	0	0	11.2	10.6
Q2 26	11	0	10.6	10.9

A quarter with eleven links did marginally worse than a quarter with none.

I want to be careful about how much that means. Two data points, ordinary volatility, and a June update in the middle. It does not show that links don't work. What it shows is that the effect I was confidently attributing to 44 links in January is smaller than the quarter-to-quarter noise on my own site, which means my original table could not have detected it either way.

What I'd say to anyone who finds this thread later: the anchor question is probably unanswerable at the scale any of us operate at. Not unanswered. Unanswerable, because the effect is smaller than the noise and we can't afford holdback groups.

I'll keep buying branded. Not because I've shown it's better. Because it's the option that can't be a mistake.

"The option that can't be a mistake" is the most honest description of branded anchors anyone has written and it should end the argument, though it won't.

Six months, four people posting numbers, one negative result from cashflowkid, one internal test from hexforge with an 8-point gap he won't claim, and a fifteen-month curve whose author has talked himself out of it. Nobody has produced a case where an external anchor moved anything.

That's not because the tactic is being suppressed. It's because it stopped being a tactic somewhere around the time the engine got good at reading the page, and what's left is a habit we argue about annually.

Someone will start this thread again in 2027 and I'll be in it.

you'll be in it saying the same thing and you'll be right again. still doesn't help anyone pick an anchor on a tuesday