Person or AI Bot? How to Spot, Stop and Use AI Auto-Replies Properly

If you've noticed that some LinkedIn messages and emails are replying a little too fast, a little too often, and at hours no normal person is at their desk, you're not imagining it. A lot of outreach is now run by AI sequences and auto-responders, and once you know what to look for, it's very obvious.

This article covers three things: what's going on, how to stop it reaching you, and if you want to use AI on your own messages, how to do it without turning into the thing everyone's annoyed by.

1. What's happening

Sales and recruitment tools now plug an AI model straight into LinkedIn and email. Someone uploads a list of names, writes a prompt along the lines of "get this person on a call", and the tool sends the first message, reads your reply, and writes the next one. Many run around the clock, which is why you get a cheerful answer at 2am on a Sunday.

The important bit is what they're told to do. Most are set up to treat anything you say, including "not interested", as an objection to work around. So the conversation never quite ends. Every reply finishes with another question, because the bot's whole job is to keep you talking until you book a meeting.

Some of these are fully automatic. Others are a real person clicking "send" on whatever the AI drafted without reading it properly. From your side it feels much the same.

How to spot an AI auto-reply

• The timing is wrong. Replies within a minute, at weekends, in the middle of the night, or follow-ups landing at suspiciously exact intervals (day 3, day 7, day 14).
• It never lets go. Whatever you say, the reply pivots back to a question. "That makes total sense! Out of curiosity, how is your team currently handling X?"
• The flattery is generic. A compliment about your "impressive journey" or "inspiring work" that could apply to anyone, often lifting a phrase straight from your headline.
• It doesn't answer you. Ask a specific question and you get a warm, vague paragraph that responds to the general topic, not to what you said.
• The tone never changes. Relentlessly upbeat and polished across every message. Real people get shorter, more casual or a bit distracted as a conversation goes on.
• The writing tics. "I completely agree", "great question", bullet points appearing out of nowhere in a chat message, and words like delve, landscape, leverage and game-changer.

One of these on its own proves nothing. Three or four together, and you can be fairly sure.

2. How to stop it

Here's the catch. Polite, vague replies like "maybe later" or "not right now" are exactly what these systems are built to handle. They read as a soft no, and soft nos get followed up. You need to be clear, not clever.

Option 1: The clear stop

Short, friendly and impossible to misread as interest:

"Hi [Name], thanks, but this isn't something I need. Please remove me from your list and stop any further follow-ups. All the best."

Good tools recognise an opt-out like this and stop the sequence. If a real person is behind it, you've been perfectly polite.

Option 2: The gentle callout

If you suspect a human is leaning on AI and you'd like a real conversation:

"Hi [Name], these replies read as if they're automated. If you'd like to talk, I'm happy to, but I only reply to messages written by a person. Otherwise, please take me off your list."

Option 3: The direct callout

If you're fed up and want to be blunt, this is a message I sent recently on LinkedIn:

"Are you using AI to generate messages automatically? I don't feel like I'm talking to a real person here. Since August 2026, the EU AI Act has required AI that talks directly to people to make it clear it's AI. You might want to stop doing this to people, it's blatantly unfair."

The AI Act has teeth. Breaking its transparency rules can mean fines of up to €15 million or 3% of a company's worldwide annual turnover, whichever is higher. Those fines are aimed at the companies building and running these tools rather than the person sending you the message, but it's a good reminder that the EU takes undisclosed bots seriously.

Option 4: Say nothing

For cold LinkedIn messages, not replying is often the best use of your time. Most sequences stop after three or four unanswered steps. Don't react with a thumbs up either, some tools log any reaction as engagement and carry on.

If it still won't stop

• On LinkedIn: use the message menu to report it as spam, or block the sender. Using bots to send messages is against LinkedIn's User Agreement, and reports are how those accounts get restricted.
• On email: mark it as spam rather than just deleting it, so your inbox learns. Gmail and Outlook both get better at filtering the more you do this.
• If you're in the UK or EU: you have a legal right to object to direct marketing under GDPR, and once you've objected, they have to stop. You don't need to explain why. A short "I object to the use of my data for direct marketing, please delete my details" does the job.

3. If you want to use AI on your own messages, do it properly

There's nothing wrong with using AI to cope with a busy inbox. I use it every day. The problem isn't AI, it's letting it talk to people on your behalf without you reading it first. Here's how to get the time savings without becoming someone else's 2am bot.

Start with triage, not replies

The biggest win isn't AI writing your messages, it's AI sorting them. Have it read what comes in and label it, something like "needs me", "can wait", "newsletter" and "sales pitch", so you only spend your attention on the messages that need a human brain. I use Claude Cowork for this each morning as part of a daily triage routine, and it's saved me far more time than auto-replies ever could.

Let it draft, but you send

Most email apps and AI assistants can now write a reply for you. Use that, but keep yourself as the last step. Read it, cut the waffle, add one specific detail only you would know, then send it. That one detail is the difference between a message that sounds like you and one that sounds like everyone else's.

Rules worth sticking to

• Never let it send on its own to real contacts. Fully automatic replies will eventually misread sarcasm, a complaint, or bad news, and answer it cheerfully.
• Respect "no". If anything you set up does outreach, make sure a clear no ends the conversation straight away. Don't programme it to "overcome objections".
• Keep working hours. If you schedule messages, send them when a person would. A reply at 2am on a Sunday tells people exactly what's going on.
• Be upfront. Since August 2026, the EU AI Act has required AI that talks directly to people to make it clear it's AI, unless that's already obvious. The legal duty mostly sits with the companies behind the tools, but the principle applies to you too, and people trust you more when you're honest about it.
• Avoid LinkedIn automation tools. Tools that send connection requests and messages for you break LinkedIn's terms, and accounts using them regularly get restricted. Your LinkedIn profile is worth more than the leads.
• Teach it your voice. Give it a few examples of messages you've written and a short list of words you'd never use. It'll still need editing, but far less.

The short version

If a message replies too fast, too often, too cheerfully and never quite answers you, it's probably a bot. Tell it clearly to stop, report it if it doesn't, and don't waste your energy being polite to software. And if you're using AI yourself, let it sort and draft, but keep the sending, and the relationship, for yourself.

If you want more practical AI workflows like this one, join my newsletter below. I share what I'm really using and testing, not just the polished results.