Every eighteen months someone announces that guest posting is dead. Every eighteen months it turns out to be one of the most reliable ways to build authority on the web, provided you are doing the version that resembles publishing rather than the version that resembles laundering.
What has genuinely changed since 2024 is the spread of outcomes. Good guest posts work about as well as they always did. Bad ones have gone from "slightly useful" to "actively worthless", and quite often to "a liability sitting in your profile".
What Google actually went after
Three separate things get conflated in this conversation, and they have different consequences.
Link spam detection got substantially better at identifying links that exist purely for ranking manipulation. The result is usually devaluation rather than penalty — the link stops counting and nobody tells you. This is the outcome most link buyers experience without ever realising it.
Site reputation abuse targeted established sites renting out sections to third-party content that has nothing to do with their actual audience. If you have seen a respectable regional newspaper hosting a "best casinos" directory, that is the pattern. Those sections got hit hard, and buyers who paid four-figure sums for placements on them lost the lot.
Helpful content signalling reduced visibility for sites producing content with no evident purpose beyond ranking. Plenty of guest post farms fell into this by construction — they publish on every topic, for no audience, with no editorial identity.
Notice what is absent from that list: publishing a well-written article on a relevant publication that links to a source. That has never been the target, because it is indistinguishable from how the web has always worked.
The test that actually predicts outcomes
We apply one question to every piece we publish, and it has held up better than any metric we have tried.
Would this article deserve to exist if the link were not in it?
If yes, you have an editorial placement. If no, you have a link with some words wrapped round it, and the words are getting thinner every year as detection improves.
This is not a moral position, it is a practical one. Articles that pass the test get read, get indexed, sometimes rank for their own long-tail terms, and occasionally earn links of their own. Articles that fail it depend entirely on Google not noticing, which is a deteriorating bet.
How to assess a site before you buy
Most of what you need is visible in ten minutes without a single SEO tool.
- Does it rank for anything other than its own name? A publication with genuine authority ranks for topic terms. If the only thing bringing traffic is the brand name, there is no authority to pass on.
- Are recent articles indexed? Search a distinctive sentence from a post published in the last month. If nothing comes back, your article will not be indexed either.
- Is there editorial coherence? A site covering crypto, weddings, dentistry and lawnmowers is not a publication with broad interests. It is inventory.
- Do the authors exist? Eleven authors with stock-photo faces and no history anywhere else is a tell.
- Would a human read this? The oldest test and still the best.
Also check the obvious commercial signals. A site with a "write for us" page listing per-post pricing by domain rating is selling inventory, not accepting contributions. That is not automatically disqualifying — plenty of legitimate publications take paid placements — but it tells you what you are actually buying.
Where the value went instead
As guest post farms lost effectiveness, three things picked up the slack.
Relevance-first placement. A modest site that genuinely covers your industry now outperforms a bigger, broader one. Topical proximity has quietly become the strongest single predictor of whether a link does anything.
Link insertions on aged pages. A link inside an article that has already been indexed, crawled repeatedly and ranked for something starts from a much stronger position than one on a page published this morning. Cheaper, too — more on that here.
Coverage-led work. Digital PR that produces brand mentions and branded search lift compounds in a way pure link acquisition does not. It is slower and harder to attribute, which is precisely why fewer people do it.
What a good guest post looks like in 2026
- 700–1,200 words with an actual point of view, not a survey of a topic
- Published on a site whose readers plausibly care about the subject
- One contextual link, placed where a reader would find it useful
- Anchor text that reads like a sentence, not a keyword
- Written for the publication's voice rather than pasted in unchanged
- Specific enough that nobody could swap your brand for a competitor's
That last point is the one most people skip. If an article would work identically with any other company's name in it, it is not really about anything — and increasingly, that is detectable.
The honest summary
Guest posting in 2026 works when it looks like publishing and fails when it looks like a transaction. The cost of doing it badly has gone up, the cost of doing it well has stayed roughly flat, and the gap between the two outcomes is now large enough that the cheap option is genuinely the expensive one.
If you would rather not audit forty sites yourself, that is what we are for — how our guest posts work, or pitch us one if you want to write it yourself.
Frequently asked
Is guest posting still effective in 2026?
Yes, but the distribution of outcomes has widened. Guest posts on relevant, genuinely maintained publications still pass value reliably. Guest posts on sites that exist only to host guest posts increasingly pass nothing at all, and the gap between the two has grown every year since 2024.
Does Google penalise guest posting?
Google devalues far more often than it penalises. The usual outcome for a bad link programme is not a manual action but silence — the links simply stop counting, and the money is gone with no notification that anything went wrong.
How can I tell if a site is worth guest posting on?
Check whether the site ranks for anything other than its own brand name, whether recent articles get indexed, whether there is a coherent editorial identity, and whether any human would plausibly read it. If a site publishes on eleven unrelated topics and ranks for none of them, it is inventory, not a publication.
Want this done rather than explained? We own 38 UK publications and place editorial articles and links on them from £65. Send the target page and we'll come back within one working day with a plan.