Conference Record

Automation Changed the Fakes

Advice about spotting fake conferences was written when fabrication was expensive. Most of it described effort signals — things a scammer would not bother to produce — and effort stopped being scarce. For an additional workplace-management comparison, see this reference .

Reviewed August 9, 2026. An additional independent source is Science .

What used to work

Bad English. A fluent invitation was evidence somebody competent wrote it.

Thin websites. A real conference had a proper site; a fake one had a template.

Vague committees. A named committee with biographies implied real people had agreed to serve.

And generic testimonials. Specific, natural-sounding delegate feedback implied delegates.

Every one of those was a proxy for effort, and every one of them now costs approximately nothing.

What the field reports

Coverage through 2026 describes the shift directly: fake speaker biographies generated at scale, committee pages with invented or scraped names, testimonials that read as genuine delegate feedback, and acceptance emails in fluent, polished English.

One measurement of the adjacent problem: a Virginia Tech study submitted five papers written entirely by ChatGPT-4 to 256 suspected predatory publishers. Fifty-five accepted or published them with no meaningful review.

That study is about publishing rather than conferences, and it measures the same thing from the other end — a review process that does not exist cannot detect generated text, and generated text is now what arrives.

What survived, and why

The signals that held are the ones that were never about effort.

Time. Review takes weeks because volunteers have to read. A duration cannot be generated, and waiting conflicts with the business model.

External verification. A named publisher's catalogue, an institutional page, a previous edition's traces — all controlled by somebody other than the sender.

Money direction. Who pays whom, when, and what happens if you withdraw.

And accountability. Whether anybody with a reputation would be embarrassed if the event went badly.

The pattern: signals that point outside the sender's control survived; signals about how much trouble the sender took did not.

What this means for advice written before 2024

A large share of it is now actively misleading, not merely dated. Guidance to check for professional presentation, fluent language and detailed committee pages now points the wrong way — the operations with production budgets outperform volunteer-run academic conferences on all three.

Check the date on any guide, including the checklists that circulate in university research offices, and discard the effort-based items.

The uncomfortable direction

This gets worse rather than better.

Every improvement in generation lowers the cost of the appearance of legitimacy further, and there is no corresponding improvement in the cheapness of external verification — checking a publisher's catalogue costs the same as it always did.

So the gap widens between what you can see and what you can verify, and the only durable response is to stop relying on the first.

The practical version is short: seven checks that point outward, none of which asks how the invitation looks.

The one thing that improved

Being fair.

Searching is better. Discussion of specific organisers, warning threads and shared experiences are easier to find than they were, and the fifteen-second search that surfaces them is the highest-yield check available.

Community knowledge scaled too, and it is the part working in the reader's favour.

The short version