Stop interviewing chatbots

It's tech PR, not journalism. Cover them as products.

Dear reporters, editors, and serious people who read their work,

When journalists are assigned, or assign themselves, to make hay of their interactions with chatbots, it’s PR and the ruse is already worn out.

Admittedly, several years ago I wrote a column about Replika, a chatbot companion that was comparatively primitive six years ago, and about its business model. The piece is pretty dated now. Read it if you wish; but if you hit a paywall, you’re not missing much. In its defense, the chatbot is never treated as a sentient being.

Chatbots are not alive. They are not conscious agents. What we are dutifully calling “artificial intelligence” at the request of its developers and investors is a suite of highly powerful software applications, some of which function autonomously.

The goal of building Artificial General Intelligence, which theoretically could think independently and even surpass human reason and creativity, is hypothetical. AI has yet to produce a conscious moral and creative actor in the physical world.

Interviewing chatbots is a gimmick and the resulting bad-faith articles teasing belief in their sentience are painful to read.

You can interview Zoltar, the coin-operated robot fortune teller or Kermit the Frog, or an empty chair standing in for a political opponent; but here the audience is in on the bit and there is no bad faith or quackery.

But interview a chatbot? Quack quack quack.

An AI séance

One repugnant example of this flimflam was Jim Acosta’s interview of a digital avatar mimicking a child who was murdered in a mass shooting in 2018.

Last year, Acosta left CNN and launched himself as a Substack producer. This was one of his early ventures on the platform.

On BlueSky, Acosta teased his story by writing, “I’ll have a one of a kind interview with Joaquin Oliver. He died in the Parkland school shooting. But his parents have created an AI version of their son for a powerful message on gun violence.”

Presenting this spectacle as “an interview with Joaquin Oliver” via AI séance was grotesquely misleading and arguably exploitative of a family’s grief, even if seven years had passed and the family actively participated in the story.

It was hideous and Acosta was roundly panned for it. A photograph of the deceased twitched in a sickly imitation of life while a digital simulation of his voice retrieved from its database some humanoid responses to questions Acosta asked it.

Yet the scathing reaction was less about the quality of the avatar than Acosta stooping for a weird jig when he might have interviewed survivors of mass shootings, instead pulling a self-serving stunt at the expense of a child who, having died seven years prior to this farce, was certainly in no position to consent to how his image and voice were being mimicked to say things on his behalf.

Taffy meets Tilly; it’s mostly rather silly

Earlier this month, the New York Times — which I had just barely started to forgive for their credulous, nonsensical apologia for Uri Geller’s spoonbending chicanery in 2023 — drew some scorn for its magazine’s entry into this troubling field.

It follows a formula as familiar as a romcom or buddy movie, in which reporters present themselves as sophisticated and urbane enough to know their subject is software-not-soul, yet find themselves guilelessly, breathlessly confuzzled that the chatbot is so convincing, so adept, that there just may be some real connection.

Heck, sometimes smart people get so high on what chatbots are supplying they conclude that the things are conscious. Case in point: Blake Lemoine, the Google engineer who decided his chatbot was sentient; or, more recently, Richard Dawkins, evolutionary biologist and pugnacious atheist, who seems to have fallen down the rabbit hole himself.

But we don’t need to make fun of them: One in three people who use chatbots appear to entertain the possibility at some point that the things are alive and that has much to do with our own wiring.

Anyway, the Times Magazine knows how to set up the buddy movie. “I profile celebrities for a living,” states the internet headline of Taffy Brodesser-Akner’s story, “Nothing prepared me for Tilly Norwood.”

Tilly Norwood, for those unfamiliar, is the name of an avatar realistically simulating a young woman that has been presented by its designers as an AI actress. The designers are with a production company founded by Eline Van der Velden, an actor and producer in her own right.

The character made its debut in a video in 2025 and the creators cleverly stoked publicity with a press release claiming there were talent agents lining up with pens poised to sign this smooth-pixeled ingenue up and launch it into a field famously low on attractive young women trying to earn a living.

With the proliferation of highly powerful computing and sophisticated imitations of human arts and interaction throughout the economy, media arts, medicine and therapy, the complexity of human interactions with this technology is irrefutably a relevant topic for journalism.

Journalists should be interviewing employers, software developers, workers who think their jobs could be threatened in the future, union leaders, psychologists, economists and ethicists about technology’s implications for humanity.

This was not that. And I’m not going after this writer personally so much as her editors. In fact, I feel like Taffy Brodesser-Akner may have gotten used here. But when the Times, as your employer, flies you to London to talk to a digital cartoon character on a laptop computer, what do you do?

And if the Times tries to rope your own reputation into the persona — bringing you into the story as a seasoned celebrity interviewer sitting down with the AI actress, subscribe to read the rest — do you object?

(Ahem. You should object. And you deserve to have a champion speaking up for you. Don’t let yourself get used that way if you can avoid it.)

Brodesser-Akner writes it up with humor and glowing embers of skepticism about AI’s promises. She describes her interactions with the chatbot as exhausting, like “the feeling of being at a computer all day,” of looking for depth and not finding any.

Yet she keeps playing out a comedy of reminding herself that “Tilly is just a computer,” which provokes me to ask What does Tilly smell like, then?

Also, since the writer describes unnatural twitches by the avatar during their interview, I don’t believe she had trouble remembering it was a computer. Not calling anyone a liar, here, just sniffing the air and detecting wafts of dramatic license.

She also interviews Van der Velden and shares her story with us, which is the highlight of the article. As a reader, I am far more curious about the woman behind the curtain, her motivations, what led her here after working as an actor herself.

“Pay no attention to that man behind the curtain.”

“Tilly” responds to questions about acting techniques with information readily available to large language models, but has never attended an acting class or confronted itself the way humans do when they step through the doors opened by acting exercises and methodical scenework.

I thought it was interesting that “Tilly” picked out the Meisner technique in particular, since the exercise of repetition for which Meisner is best known has a parallel in a therapeutic technique chatbots have successfully imitated.

What was disappointing is the absence of any questions about how “Tilly” is a composite of human artists’ work, the faces, voices and dramatic performances that have been vacuumed, absent consent or contract, into the chatbot’s data bank.

Take it from me, an acting teacher long before I was a journalist: Imitation is a skill, but it’s not acting. The creative acts that a seasoned performer practices and executes are, as yet, beyond the capacities of a chatbot. And I’m not saying that to curse chatbots, as much as I love cursing them.

The New York Times Magazine story took no interest in this. It was framed by the writer and editors, then marketed by the publisher, as a celebrity profile.

A tongue-in-cheek PR story is still a PR story and this, functionally, was that.

Cover chatbots as products

For a story about how employers are using chatbots to screen job applicants, I’ll grant reporters a variance if they want to document the experience. This, after all, might be a story showing how AI is being used in the workplace — how the product works.

But keep it clear that the chatbot is a product and report on its uses, its outcomes, its wastes and how well it fulfills its producer’s promises.

Indulging the fantasy of consciousness, by a professional news organization, is akin to endorsing other kinds of woo.

Like that baseless and debunked fantasy that Uri Geller has paranormal powers. Did I mention I’m still pissed off at the Times for that inane story?

Keep your wits about you, everyone! Fend off the bullshit and écrasez l’infame.

Warmly,

A