The field, measured

I reviewed 614 speaker websites. Nine had nothing wrong with them.

Not an opinion survey and not a roundup of other people’s articles. Every site below was opened, screenshotted and checked against the same short list of faults, the kind an event organiser hits in the first few seconds without looking for them.

9of 535 were clean

cleancarried a faultone mark per site

What was wrong

Nothing exotic. The same seven things, over and over.

Percentages are of the 535 sites that could be assessed. Each site is counted once, under the single fault recorded against it, so these seven do not overlap: together they are 386 of the 535. Another 117 carried something milder, 23 something too rare to list, and 9 nothing at all. Bars run to 16%, not to the largest of them.

A layout collision
The menu sitting on top of the logo, a headline running off the edge of the screen, a face cut in half by the header.
15.3%82 sites
Broken outright
The address no longer opens, the browser stops you with a warning first, or the domain lapsed and now shows a for-sale page.
12.9%69 sites
Text you have to work to read
Copy set on a photograph or in a grey too light to hold at arm’s length.
11.4%61 sites
Visibly of another decade
The weakest category here, because it is a judgement rather than a fact.
10.7%57 sites
Something covering the content
A cookie banner, newsletter popup or chat widget sitting on the headline or the only booking button.
9.3%50 sites
No headline at all
Nothing on the first screen says who this person speaks to, or what changes when they do.
8.6%46 sites
Nowhere obvious to enquire
Nothing to click to get in touch before you scroll, on the page an organiser lands on first.
3.9%21 sites

– the thing everyone gets wrong about old websites.

“Dated” is not the problem. It is barely even measurable.

The obvious way to find a bad speaker website is to look for an old one: page code from the 1990s, nothing telling a phone how to size itself, fonts from 2009. Across the sample those markers almost never appear, because everyone is on Squarespace, Wix, WordPress or a builder that turns out clean modern code no matter how badly the page in front of you actually reads.

1.9%: were laid out with tables, the 1990s way to build a page
13.3%: never told a phone how to size the page

Age was the wrong question. What recurs is faults, and a fault is a fact rather than a matter of taste.

And underneath, quietly

15.7%can be opened on an address the browser marks not secure
3.1%fail if a visitor types www in front of the name
0.9%stop a visitor with a full-page security warning

Checked on 350 of the sites by connecting to each one, not guessed from the page. Bars run to 20%.

How it was done

Method, and what this does not tell you

9cleannothing wrong with them526at least one faultof the sites that could be assessed79excludedblocked the check, failed to load, or were the wrong person

614 unique sites, gathered from speaker bureau listings, association directories and conference pages across the UK, Ireland, the Netherlands, the Nordics, Australia, New Zealand, Canada and the US. Each homepage was opened, screenshotted and reviewed by eye against the same fixed list of faults. The plumbing was checked separately: whether the address still resolves, whether the security certificate is valid, and whether an insecure address forwards to a secure one. Last computed 24 August 2026.

79 sites are excluded from every percentage. Sixty-eight blocked the automated check or failed to load. Eleven turned out to be the wrong person: the address was a guess at a name, and the page that opened belonged to a namesake rather than to a speaker. Counting either kind as clean would have flattered the result. That leaves 535.

This does not measure whether a site works. It cannot see whether the talks are named, whether there is a reel, or whether a fee is stated anywhere. Those are the questions in the thirty-second test, and answering them needs a person reading the page, not an automated check. Nothing here should be read as a claim about anyone’s business.

No individual site is named here and none will be. The figures are aggregate, the underlying list is not published, it is not in this site’s source code, and the drawings above are of a kind of fault rather than of anyone’s page.