smarticulate

Benchmark

How UK agencies sound

Twenty-two agencies, six pages each. The best sector measured, and nobody at the top.

Twenty-two UK brand, marketing and content agencies, read by the same code on the same day, six pages each. The people who sell voice for a living have one, and score higher than any sector measured so far. None of them reaches the top band, and the reason is the same one as everybody else's.

What was measured, and what could not be

Twenty-two of forty-five, and why the number is small

Forty-five agencies were listed, from the independents to the network shops. Seventeen could not be fetched at all, which is the highest failure rate of any sector: agency sites are built on the newest stacks, and a page that is a script waiting to draw itself is a page with no prose in it when a reader arrives without a browser. Three more had too little continuous prose across six pages to score, one address had gone, and two failed without saying why. Twenty-two remain. Enough for a median, not enough for a fine one. This benchmark will be rerun with corrected addresses before it is quoted as settled.

Each was read the way a Wordprint reads a customer: the homepage and up to five of its own pages as one body of prose, 2,800 words for the median agency. The published method scores four things and reports eighteen more. No agency is named and no page text was kept.

The sector that cannot be read is a finding too.

The headline

A median of 81, the highest measured

The sector averages 76.3 and the median agency scores 80.6. The bottom tenth is at 56, the top tenth clears 87, and the best agency measured reaches 89. Eighteen of the twenty-two are in the seventies and eighties, four are in the middle band, and none is below 40. Accountancy on both sides of the Atlantic sits around 72; the AI consultancies at 56.

Three things put agencies at the top. Rhythm first. A median sentence-length variance of 9.7, the widest of any sector, on sentences averaging fifteen words. Ownership of vocabulary: only one phrasebook term is used by a fifth of them, and eight of the twenty-two cross the sixty-per-ten-thousand line against all forty of the AI consultancies. And the pages are prose: 74% of the words on a median agency site sit in continuous sentences, more than the accountants and far more than the consultancies. Agencies also write the most relaxed English of the five sectors, with contractions at 201 per ten thousand words, and they are the only sector in which "I" appears at all on more than a handful of sites.

What they do not do is cross 90. Nobody is established. The ceiling is 89.

The failure

Nothing lands: twelve agencies out of twenty-two

The same fault as the accountants. This sector should know better. Landing is the share of sentences at five words or fewer. Real voices run between 15% and 29%. The median agency manages 6.5%; twelve of twenty-two are under 8%. The top tenth lands one sentence in five, which is the figure the whole sector could reach, and the difference between an agency at 80 and one at 89 is almost entirely that.

It is an odd fault for the trade to have, because a short sentence after a long one is the oldest move in copywriting, and every agency here can do it for a client. On their own pages the paragraphs are handsome and even. Nothing in them stops. The writer who would cut a client's sentence to four words has not been allowed near the agency's own about page.

The cobbler's children, measured.

The shared phrasebook

One word

These are the phrases used by a fifth or more of the agencies measured. In accountancy the list ran to eleven entries; among the AI consultancies to fifteen and beyond. Here it is one.

PhraseFirms using it
unlock9 of 22

That is the strongest result in the five benchmarks. Agencies do not write from a shared template. A buyer reading three of them will meet three vocabularies, and whatever makes any two of them hard to tell apart, it is not the words.

Who is speaking

The agency, on sixteen sites out of twenty-two

The agency outnumbers the client on sixteen sites: "we" and "our" at a median of 509 per ten thousand words, "you" and "your" at 279. Only the recruiters say "we" more, which is a thing to notice in a trade that advises clients to talk about the reader. Four agencies let a person speak as "I". Long dashes appear on twelve of the twenty-two at a median under three per ten thousand words, which is British business English behaving normally and says nothing about authorship.

What the top decile does

Two habits that separate 87 from 81

  • It lands. The top tenth stops one sentence in five. The median agency stops one in fifteen. Nothing else on the instrument separates the top of this sector from its middle by as much.
  • It talks to the client more than about itself. The agencies at the top run "you" closer to "we" than the rest; the ones in the middle band are the ones where the agency does all the talking.

Neither needs a rebrand. Both are things every agency here charges clients for.

Limits

What this benchmark is not

Not a ranking, and no agency is named here or in the data behind it, which stores numbers only. Not a judgement on the work: an agency at 56 here may run the best creative department in the country. Not an AI detector. And it is the thinnest sample of the five, for the reason given at the top, so treat the median as a first reading rather than a settled one.

Measured 9 September 2026. Twenty-two agencies, six pages each, numbers only.

Where does your agency sit?

Paste a page you have published into the free test, choose the sector, and get the same four measures beside the sector median in about a minute. If you would rather put a client's site against the three firms it loses to, that is a Soundalike. And if you would rather sell the specification inside your own work, that is the partner arrangement.

Ian Sharpe, out thinking in a field
Ian Sharpe

Written by Ian, with Claude assisting, and scored by the smarticulate method at 99 out of 100 before it went up. Ian spent twenty-odd years turning thinking, his own and other people's, into something a board would act on, and built smarticulate to put numbers on it. His practice is ConsultingPartner.io: enterprise experience, right-sized. Who is he? →

More guides: the UK accountancy benchmark, the AI consultancy benchmark, a brand voice guide template.