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

Everyone on this board says branded anchors only. Nobody posts what happened when they did it, so here is mine, warts included.

Setup: one site, mid-competition vertical, about 180 pages. From April last year I bought 44 links across nine months and every single anchor was either the brand name, the bare domain, or a naked URL. No exact match, no partial, no "best X for Y". Before April the profile was roughly 30% exact match built over two prior years.

Quarter	Links bought	Anchor type	Target group median position, start	End
Q2	14	branded / bare	18.4	16.1
Q3	17	branded / bare	16.1	14.9
Q4	13	branded / bare	14.9	11.2

So it moved. Nine months, 44 links, median position on 26 tracked commercial phrases from 18.4 to 11.2.

What I can't tell you is the counterfactual. There were three core updates in that window and the site also gained about 40 pages of content. I have no holdback group because I only have one site in this vertical and I wasn't willing to run half of it deliberately worse.

The reason I'm posting anyway is that the entire board's anchor advice appears to rest on nothing published since about 2019. If someone has newer numbers, post them and I'll shut up.

44 links and 40 pages of content in the same window. you measured two things

I'll date the whole argument because nobody ever does and then people wonder why the advice contradicts itself.

Exact match anchors were the strongest ranking lever that has ever existed, roughly 2007 to 2012. You could rank a page for a phrase by pointing forty links at it with that phrase on them and almost nothing else mattered. Penguin in April 2012 ended it as a free lever and made it a risk. From 2012 to about 2016 the game was ratio management — people published charts of the ideal percentages and sold spreadsheets.

It has been dead since about 2018, which is exactly why "branded only" became the default advice. So your table is real and it is not measuring anchors. It's measuring 44 links. You'd have got a similar curve with exact match, and probably a slightly better one in the first quarter and a slightly worse one after the third core update.

I'm going to disagree with the strong version of that, and I'll declare that I sell links so my incentive is to say the anchor matters. It's in the signature.

What I see across a lot of client profiles is that anchors still do something, but the thing they do is narrow. They're not a lever on a competitive head term any more, agreed. They still appear to move pages for phrases nobody else is actively targeting — long, specific, boring stuff where the page has no other strong signal about what it's for.

Depends what you're running. If you're fighting for a two-word commercial phrase, old_growth_seo is right and your anchor is decoration. If you're trying to get one of 400 programmatic pages to rank for its own specific thing, the anchor still seems to be part of how it gets sorted out.

No holdback, three core updates, 40 new pages. What is the table evidence of, in one sentence?

That a site doing branded-only anchors for nine months did not stall, which is a weaker claim than most people on this board make in either direction.

I'd take the criticism further than you have, actually. The Q4 jump is the biggest of the three and it's the quarter with the fewest links and the most content. If you read the table naively it argues against my own thesis.

I'm not claiming causation. I'm claiming a published number in a conversation that has been running on assertion for six years.

this is great, saving it. i've been branded-only since last summer because everyone said to and i've never once checked if it did anything

might actually run exact match on one page and see what happens, someone should

This thread is about to spend forty posts on external anchors while ignoring the anchor surface that is a hundred times larger and entirely under your control, so let me lay that out and then come back to the actual question.

The proportions nobody looks at

MetricsOverMagic bought 44 external links in nine months. His site has 180 pages. If those pages link to each other at even a modest density, he has somewhere between 1,500 and 4,000 internal anchors pointing at his own URLs, all of which he wrote, all of which he can change this afternoon, none of which carry any risk at all.

The external anchor debate is an argument about 44 things while 3,000 sit unexamined.

What I think the evidence actually supports

1. External exact match: a strong lever until 2012, a managed risk until roughly 2018, largely devalued now. I agree with old_growth_seo on the direction and I'd put less confidence on the endpoint than he does.
2. External branded: does not appear to hurt, does not appear to target. Its function is that the link exists.
3. Internal anchors: still visibly effective in every test I have run, including three last year. Change the internal anchor text pointing at a page from generic to specific and the page's impression profile shifts within two to five weeks. This is the most reliably reproducible thing in link building right now and almost nobody talks about it because you can't sell it.

What I'd want from a real test

MetricsOverMagic's design problem is that he has one site. The test that would settle the external question needs:

• Two groups of target pages on the same site, matched on current position and traffic.
• Identical link volume to both, differing only in anchor.
• A window with no content changes to either group.
• Twelve weeks minimum, ideally spanning no core update, which in practice is impossible now.

That is a genuinely expensive experiment and the reason nobody has run it is that the person with the money to run it has no reason to publish the answer.

Where I'm uncertain

I don't know whether the internal anchor effect I keep seeing is the anchor or the fact that changing internal anchors usually means changing which pages link to which, and the second is a structural change. I have not designed a version that separates them and I've been meaning to for two years.

Summary: your external anchor choice is probably close to irrelevant and your internal ones are probably not, and the board has this exactly backwards.

Six years of arguing about the forty-four.

internal linking?? nobody has ever once mentioned this to me

IME the internal one is also the only part of this anyone can test properly, because you own both ends.

Split your pages into two matched groups. Rewrite internal anchors for group A only, change nothing else, and diff impressions per page against group B over eight weeks. No third party, no purchase, no waiting on a broker. You can run it this week.

The confound hexforge flags is real but you can control it: keep the same set of source pages linking to the target and change only the text in the anchor. Same links, same structure, different words. That isolates it.

I'd do it on 40 pages minimum. Below that the impression noise eats the effect.

This might be a stupid question but when you say internal anchor, do you mean the text in the body links, or does the navigation menu count too? Mine says the same thing on every page and I've never thought of that as an anchor at all.

it counts and it's the same words on 180 pages, which is why it does nothing

I've got a problem with point 3 and it's the same problem I've had with internal linking advice since about 2016.

Every time somebody reports that changing internal anchors moved a page, what they've actually done is gone through a site and thought hard about what each page is for. The anchors changed. So did the surrounding sentences, the choice of which page to link from, and often the page title, because you can't rewrite an anchor without reading the paragraph it's in.

What you're measuring is attention. The anchor is the artefact of the attention, not the mechanism.

I'm not saying don't do it. I did it to a client site in 2019 and it worked and I billed for it. I'm saying that "change internal anchors" as an isolated instruction, handed to someone who does it mechanically with a find-and-replace, has never produced a result I've seen. It only ever works when a person reads the site.

PipelineRat's design would catch this, incidentally, if the anchors were changed mechanically. Which is why I'd want it run both ways.

That's a better statement of my own uncertainty than I managed and I'll take it.

Of the three tests I ran last year, two involved me reading the site and rewriting properly. The third was closer to mechanical — a template change that altered the anchor pattern across a section — and its effect was much weaker. I noted that at the time as "smaller sample" and old_growth_seo's reading is at least as good an explanation.

So I'd revise point 3 to: reading your site and fixing how its pages refer to each other works, and I do not know how much of that is the anchor text specifically. Which is considerably less useful as advice and considerably more likely to be true.