Sherry Turkle's Artificial Intimacy, Read by a Guy Who Manufactures the Warmth

Sherry Turkle's Artificial Intimacy came out September 29. It isn't about marketing. That's exactly why every brand person should read it.

The argument fits on a bar napkin. Machines can perform care convincingly now. Performed care is not care. We're starting to accept the performance anyway, and the accepting is the damage.

I have a dog in this fight. I run an AI content operation for food, beverage, CPG, and lifestyle brands. My machines render hero images through an API, grade them to a client color profile, write captions, cut video, and push the whole batch out on a schedule. The warmth you read in a brand caption at seven in the morning didn't come from a person who woke up feeling warm. It came from a pipeline I built.

Turkle wrote 288 pages of warning about what that does to people. I read it as a man reading his own job description written by a stranger who doesn't like him much.

The crowd here is tiny, and that is part of the story

Twenty-six ratings at 4.31 is a dinner party, not a market.

I run this number for clients every week. A launch with twenty-six reviews is twenty-six opinions from people who raised a hand. They're never a random draw. Thirteen left written reviews.

Several of those reviewers say they got the book free through a Goodreads giveaway or a NetGalley copy. So the earliest readers of a book about manufactured attachment came straight out of a publisher's attachment-manufacturing program. Nobody planned that joke. Still a good one.

By Christmas this book will collect thousands of ratings and the average will slide. That's what happens when a self-selected early group hands a book over to everybody else.

Read the dinner party. Watch what they disagree about.

What they actually fight about

The split on Turkle is clean. It has nothing to do with chatbots being creepy. Everybody already agrees they're creepy. The fight is whether she owes you a way out.

One reader on Goodreads loved the whole thing and wanted more of it, writing "more books against ai please and thank you." Another called the examples extreme and illustrative in the same breath, then landed here: "What we're getting with AI right now is infantilizing, something that destroys human capability and agency." A third sounded like a teacher and gave the most generous read in the pile, that "relationships and learning take time, patience, empathy, and determination."

The pushback is better. One Goodreads reader wrote that Turkle "is clearly anti-AI and doesn't provide balance for cases where AI can be incorporated," and complained she "presented case study after case study without taking the time to interpret the studies." Another said flatly, "I do wish that the author were more balanced in pointing out some of the positives of AI," then asked the question the whole book trips over: "Is talking to an AI chatbot worse than not talking to anyone?"

A third liked the book, rounded up to four stars, and still filed the complaint every editor dreads: "It feels like it could have been a longform article." A fourth wanted the practical half that never arrives, noting the book "doesn't really push hard on practical ways to keep real human intimacy strong." And one reader with an early academic copy delivered the sentence that stuck with me longest: "We will never sit in a smoky bar with a chatbot" and fall in love.

Turkle's own line, quoted by a reader who clearly underlined it, is the thesis in one breath. "Chatbots provide mirrors rather than relationships."

Mirrors. Hold onto that. It's the most useful sentence written about brand social media in a decade, and Turkle wasn't thinking about brand social media when she wrote it.

I build mirrors for a living

A brand wants to feel present to a few thousand people it'll never meet. No human can do that job for real, so it buys a system that bounces each of those people back at themselves in the brand's colors.

Your comment gets a reply that sounds written for you. The salsa photo looks like the kitchen you wish you had. The caption carries a friend's cadence.

None of that is fraud. It isn't a relationship either. The readers arguing in Turkle's threads already put their finger on why.

The one who asked whether a chatbot beats nobody asked the honest version of the brand question. Is a warm automated reply worse than silence? For a tire company at eleven at night, probably not. For a small food brand whose whole pitch is a family makes the thing by hand, that automated warmth is a slow leak in the only tire you've got.

I learned that line the expensive way, not from a book. The machine is excellent at volume, consistency, and never getting tired. It's terrible at what customers actually remember, which is almost always friction. The order that went wrong and got fixed by a person who said something unscripted. The reply that admitted the recipe was off.

Out here the sun bleaches a sign to nothing in two summers. The places people still drive to are the ones where somebody comes out and talks to you in the parking lot. No amount of rendering gets you that.

Where she overreaches, and where the critics are wrong

The three-star reviews got this part right. Turkle diagnoses the problem, stacks the cases, then tells you to reclaim the real without saying how you pay for it. That's a real gap and it earned the hits.

Readers asking for balance are making the same error I watch in brand rooms constantly. They treat balance like a synonym for correctness. A book arguing a thing is dangerous does not owe that thing a defense attorney.

Turkle spent fifty years on human-machine relationships and she finally stopped hedging. The New York Times said she traded her old wariness for full-on alarm. I'd call that a decision.

Where she overreaches is scope. She treats every machine that talks as the same machine. A grieving person uploading a dead parent's texts into an avatar and a food brand scheduling a Tuesday caption are not the same act. Flattening them costs her the readers who do this work and know the difference.

One reviewer wanted more theory and less anecdote. I wanted a taxonomy. The chapter this book needed and never got is the one that separates machines replacing a person from machines doing work no person was ever going to do.

What a brand should take from it

Two things. Neither one is an AI rule.

Stop putting "we" in captions when there is no we. That gap gets found. It gets found when a customer is already mad.

Spend your human minutes where the friction lives. Complaints. Mistakes. The one strange question in the DMs. Let the machine handle the post about the new label. That is a budget decision. Pretend it is not and you end up with a feed full of mirrors and nobody in the parking lot.

Artificial Intimacy is short, repetitive through the middle, and dead right about the thing it gets right. Four stars from me. That is roughly where its 26 readers landed before the rest of the world gets a vote.

How this piece was made

I pulled every reader quote above from the live Goodreads listing on October 1, 2026.

I read the book and built the argument myself.

It seemed dishonest to publish a piece about manufactured warmth without saying how this one got manufactured.

I ran the prose through a language model that smooths my drafts toward the way I actually talk. I locked every quoted passage so the model couldn't touch a reader's words.

An API rendered the hero images. I color-graded them to my own profile.