What AI Shortlisting Means for Your Speaker Lineup

Research from Freeman presented at IMEX America in October 2026 found 82 percent of event-goers aged 23 to 46 use AI to shortlist professional events. An assistant often reads a speaker announcement before a person does, and a lineup published as an image, a PDF or a bare list of names does not survive being summarized.

IMEX America ran from 13 to 15 October 2026 at Mandalay Bay, its fifteenth and largest edition, with more than 17,800 event professionals and over 96,000 scheduled meetings. The dominant theme was not a venue trend or a budget number. It was that software has started showing up where people used to be.

Two findings from that week matter for anyone booking speakers.

The number

Research from Freeman, presented at the show, found that 82 percent of event-goers aged 23 to 46 use AI to shortlist professional events. The same research found 75 percent treat skills-building as pivotal to whether they attend.

Ken Holsinger, Freeman's EVP of industry research, described what this does to the registration curve in a line worth keeping: it is not late registration, it is late decision-making.

That distinction has teeth. Late registration is a timing problem you solve with deadlines and discounts. Late decision-making is an information problem. People are waiting because they are still gathering, and increasingly the gathering is being done by an assistant on their behalf.

The second finding

Jeanne Sheehy, CMO at Bostrom, told an IMEX session that your next registrant may not be human, describing AI agents already interacting with event websites, registration platforms and travel booking.

Her advice to organizers was operational and sensible: treat agents like interns. Map your data, clean out stale records, give an agent only the access it needs, never administrative rights, and keep tiered oversight so that low-risk actions get spot checks while consequential decisions stay with people.

That is a systems problem for whoever runs your registration stack. The part that belongs to whoever books the speakers is different, and simpler.

What this does to a speaker announcement

If an assistant reads your event page before a person does, then whatever is not in words does not exist.

The most common way event programs are announced is the least machine-legible format available: a designed graphic posted to social, a PDF agenda, or a carousel of headshots with names underneath. A human skims that in two seconds and gets the gist. A summarizing assistant gets a filename.

The practical version, in order of how much it matters.

Put the lineup in text on a page. Not only in the image. The image can stay; it should not be the only copy of the information.

Give each speaker a reason, not just a name. A name alone is only useful to someone who already knows it. A line saying what the person has actually done, and what the session covers, is what gets summarized into somebody's shortlist.

Name the subject in the words people use. A session titled Fireside with Dana is unsearchable and unsummarizable. The same session described as a conversation on rebuilding a sales organization after a merger is both.

Say who it is for. Seniority, function, and what someone should already know. Assistants shortlisting on a person's behalf are matching against that person's situation, and a page that never says who the session suits cannot be matched to anyone.

Publish the agenda where it can be reached. A program behind a registration wall cannot be read by the thing deciding whether to register.

A first-party note, offered with its limits

This site has a small piece of evidence pointing the same way.

Across our own analytics, AI assistants are the only channel that has ever produced an inquiry here. Search sends more visitors; assistants send the ones who ask about booking a speaker. The sample is small enough that it proves nothing on its own, and it is reported here as an observation rather than a finding. But it is consistent with the Freeman number, and it was visible in our own data before that number was published.

What this is not

It is not an argument that AI decides who gets booked, and the pitch that it does is being made loudly by people selling tools.

Nothing in the IMEX research suggests assistants are choosing speakers. They are helping people decide whether an event is worth their week. The speaker lineup is one input into that, which is precisely why it needs to be readable.

It is also not a substitute for programming. A legible announcement of speakers nobody wants to see will be summarized accurately and ignored. Machine-readability changes whether a good lineup is found. It cannot rescue a weak one.

The underlying shift

For twenty years the job of an event page was to persuade a person who had already arrived on it. That job has not gone away, but a second one has appeared underneath it: to be accurately describable by something that will never see the design.

Those two jobs mostly agree. Specific beats vague for both. The subject named plainly beats the clever title for both. The main difference is that a human forgives an image and an assistant cannot read one.

Tell us who the audience is and what the session has to achieve, and we will shortlist speakers who fit, along with the plain description of each that you can publish without rewriting.