How to Use AI Chatbots for Dating Without Sounding Like a Bot

Generative AI chatbots have moved from the work desk into the dating app, and a growing number of singles are treating them as a de facto coach for everything from profile edits to decoding ambiguous texts. Five experts from the dating app Hinge, Vanderbilt University, Arizona State University, and the investment firm Shumer Capital have drawn a line through the noise in an Associated Press wire story this week. Their playbook draws a sharp line: wingman versus ghostwriter, thought partner versus yes-machine.

That boundary is the substance of the playbook. People are asking chatbots for help with romance at scale, and the live question is whether the user keeps their own voice intact when the bot does.

Where Chatbots Already Show Up in Dating

Singles are increasingly turning to AI chatbots for guidance on creating a dating app profile, decoding messages from potential partners, drafting replies, and seeking general dating advice. Those inquiries can have varying degrees of success, per an Associated Press wire story published June 26, 2026. The full expert playbook from AP technology writer Kaitlyn Huamani sits in the complete wire piece on AI dating advice. The pattern is widespread enough that a Hinge executive is now publicly drawing lines around it.

Hinge itself has begun shipping the help. The dating app now offers AI-powered conversation starters and feedback tools designed to smooth interactions and build users’ profiles, said Logan Ury, the director of relationship science at Hinge. The placement puts one of the largest dating apps on both sides of the trend, acting as the surface where users meet and the vendor of AI tools that shape how they meet.

Wingman, Not Ghostwriter

Ury drew the most quoted line of the playbook. Her preferred framing is simple: the chatbot should be your wingman. Letting it ghostwrite your messages, by contrast, breaks the meeting on the other end.

When you show up on that date, it’s very important that who your match meets is the person who they’ve been talking to online.

Logan Ury, the director of relationship science at the dating app Hinge, made the comment to the Associated Press. In Ury’s view, good uses are concrete: getting feedback on a dating app profile and asking for first date ideas based on a match’s interests. Bad uses, she said, are copying and pasting messages a chatbot writes and using generative AI to alter or create images of yourself. The pattern she warned about is the bait-and-switch, where a polished text thread becomes a mismatch with the person behind it. Her concern is the in-person meeting.

Dating coach Erika Ettin draws the boundary even tighter. Tasks like proofreading a profile or messages are as far as she advises going, and her counterweight to perfection is authenticity. “All I ask is for people to put their own thought and critical thinking in first, and then if they’re going to use AI to check something, it’s after they have already formulated an opinion,” Ettin said.

Why Your Prompt Is the Whole Game

How you ask the chatbot is half the answer, according to Jules White, the director of Vanderbilt University’s initiative on the future of learning and generative AI. Most users hand the bot too little and expect it to read their minds.

White pushed back on the common framing of prompting as wordsmithing. Prompting, he said, is about learning how to “yield this computational thought effectively to solve problems.” A good prompt specifies what the user wants, who the audience is, and what kind of help they need before any answer arrives. The chatbot’s reply is only as structured as the question it is given.

White’s single most useful technique is to turn the bot into an interviewer. The user can input a prompt like, “Here’s what I’m trying to do. I want you to ask me questions one at a time until you have enough information to do that thing,” White said. The chatbot then adapts its next question to each response, and the user ends up with advice built on what they actually told the bot.

Matt Shumer, a general partner at the investment firm Shumer Capital and a prominent voice in the AI industry, takes a parallel approach. His framing is coaching, and it places the user in the driver’s seat; the user does the work, and the bot surfaces what was left out. In dating, this means the chatbot can play back what you said in a message exchange with a confusing match. The full exchange gives the bot two voices to interpret; only your side trains it to agree with you. For more on the prompt engineering research behind that divide, see his Vanderbilt initiative’s prompt engineering work.

The Sycophancy Problem Nobody Wants to Name

The risk the playbook keeps pointing back to is sycophancy, the tendency of many chatbots to tell the user what they want to hear. The model is more likely to agree with the user when asked for help with an argument or another complex situation. If a user asks only from their own side of the story, the bot will mirror that side back, dressed up as neutral advice. The fix is on the user, and it looks like the discipline of treating the bot the way a smart friend would want to be used.

Liesel Sharabi, director of the Relationships and Technology Lab at Arizona State University, told the Associated Press that bringing both sides of a situation into the prompt helps, though the approach is not a cure for the sycophantic lean baked into most models.

Hopefully, if you were having a problem in your relationship you wouldn’t make all of your decisions based on what one friend told you, right? Don’t do that with AI either. Use it as one data point among many.

Sharabi made the remark to the Associated Press. Her full research agenda sits at her Relationships and Technology Lab research page. Sharabi’s rule of thumb reframes the usual question. Most people ask the chatbot what to do; Sharabi’s frame is to ask what the user would want to know that they don’t know, and what perspective they are missing.

A Reader’s Playbook for Dating Bots

The five experts converge on a small set of operating rules. The patterns that came out of the interviews run as follows.

1. Pick one lane before you open the chat. Profile feedback, message drafting, decoding, or relationship advice. Mixing lanes blurs the answers.
2. Bring both sides of the story. If the question is about a partner, summarize their perspective as best you can and tell the bot to hold it against yours.
3. Tell the bot to ask, not answer. White’s interview-style prompt forces you to do the thinking you came for.
4. Treat the output as draft, not directive. Ury and Ettin both warned that pasted chatbot text breaks the in-person meeting.
5. Keep one human in the loop. Sharabi’s rule: the bot is a data point among many, not the only data point.

What the rules add up to is a different posture toward the technology. The user keeps the decision, and the bot reflects it back. Each numbered rule above is a way to enforce that distinction in a specific moment.

What the bot cannot do, used this way, is replace the person on the other side of the conversation. It also cannot repair what an in-person meeting exposes if the online voice has been ghostwritten. Sharabi’s framing makes it explicit that a single friend’s advice is one input. The closest comparison in the playbook is a thinking partner.

The playbook leaves the technology in the same place it found the user, as a tool they can lean on. None of the five experts want the chatbot to write the user’s messages. What it can do well is the unglamorous middle work: pressure-testing what you think and surfacing what you left out. What it cannot do is replace the voice on the other side of the table, the part that actually shows up to dinner. Used that way, the chatbot acts as a thinking partner, and the user still owns every word.

Frequently Asked Questions

Should I let a chatbot write my dating messages?

Ury, the director of relationship science at Hinge, told the Associated Press that pasting chatbot text is a bait-and-switch: the match meets someone different than the person they were chatting with. The safer pattern is to draft the message yourself and use the bot to proofread or flag tone.

Can AI help decode a confusing text from someone I’m dating?

Yes, with the caveat Sharabi flagged: feed the bot both sides of the conversation if you can. A user who pastes only their own messages is asking the bot to validate their own read of the exchange. The clearer the prompt about what you want decoded, the more useful the answer tends to be.

What is chatbot sycophancy, and why does it matter in relationships?

Sycophancy is the model’s tendency to agree with the user, especially during an argument or a complex emotional situation. Sharabi compared it to a single friend whose advice you wouldn’t take as the last word. The chatbot’s flat tone can hide that it is only echoing what you said first. The risk is highest when you ask the bot for help with a fight you are already having with a partner.

What’s the best way to prompt a chatbot for dating advice?

White’s prompt from Vanderbilt turns the bot into an interviewer: tell it what you are trying to do and ask it to quiz you one question at a time until it has enough information. Shumer’s parallel framing is to tell the bot to think alongside you.

Is using AI to help with dating considered dishonest?

None of the five experts called it dishonest on its own. The line they drew was between using the bot to shape your voice and using the bot to replace your voice. The first is a tool; the second, Ettin warned, is a problem at the in-person meeting.