Why You Shouldn't Target Broadly in ChatGPT Ads

Broad targeting in ChatGPT Ads sounds tempting: 'Let's show ourselves to everyone, the algorithm will find customers.' However, in chat, this is a surefire way to waste your budget. Chat (a conversation with an assistant) operates differently than a search engine. I’ll show you three intent filters without coding and one metric that can help you manage your budget effectively.

Chat is Not a Search Engine: Where Intent Gets Lost

A search engine is a moment with a clearly defined goal. Someone types a phrase and expects a list of results. Chat, on the other hand, is a conversation. People often mix several topics in one thread and frequently change direction. Ads can slip in 'between the lines' when users aren’t ready to buy yet.

For example, someone asks for 'ideas for a language school newsletter' but sees an ad about implementing a CRM (Customer Relationship Management system). The intent doesn’t match. A click might happen, and the conversation may start, but it won't lead to a sale.

Conclusion: in chat, it’s not about 'the broader, the better,' but rather 'hit the intent.' Intent filters are necessary.

Three Intent Filters Without Coding

Intent is the reason someone is chatting with the assistant. Once you capture it, your budget works effectively. Below are three filters you can set up using simple messages in your ads and chat scenarios — no programming required.

• Topic exclusions: clearly state who the offer is NOT for. For example: 'Not for DIY projects, thesis work, job offers, or free tools.' If the tool allows for exclusions, use them. If not, include this in your message.
• Task matching: describe specifically what task you help with. Instead of saying 'marketing automation,' say: 'We schedule more appointments from the calendar' or 'We recover abandoned carts in e-commerce.' Name the task,
• not just the concept.
• Qualification gate in chat: start the conversation with a brief selection process. It’s like asking at a reception: 'What can I help you with?' This saves time for both parties and directs mismatched individuals to the
• educational materials instead of dragging out a long, unproductive conversation.

Qualification Gate in Chat: A 30-Second Script

The welcome script (a prompt — a short command for the AI assistant) guides the conversation. Set the goal: 'first qualification, then advice.' The assistant greets the user, asks 3-4 short questions, and only then suggests the next step: a consultation, a preliminary quote, or a link to materials. The whole process should take about 30 seconds.

Example questions: what exactly do you need help with; when do you want to start, and what is your approximate budget; what size is your company and what industry are you in? If someone doesn’t fit (for example, they are only looking for free solutions), thank them and provide a link to a guide. You build trust by clearly setting boundaries and directing them to real help instead of dragging out a fruitless conversation.

One Metric Instead of CPC: Cost of Qualified Conversation

CPC stands for cost per click. In chat, a click rarely indicates a real intention to buy. A simpler and fairer metric is the cost of a qualified conversation (how much you pay for a dialogue that meets your criteria). You can define the criteria as: matching task, matching company profile, and a realistic start date.

Formula: cost of qualified conversation = campaign expenses / number of conversations that passed the gate. For example, you spend 2000 PLN. You have 80 conversations, of which 20 are qualified. You pay 100 PLN for a qualified conversation. After implementing filters, you have 40 conversations and 25 qualified — you pay 80 PLN and talk to the right people.

• Set 3 qualification criteria (task, profile, timeline/budget). Write them in one sentence so the team understands the same thing.
• Collect data simply: a spreadsheet or CRM (a system for managing contacts). Mark 'YES/NO' for each criterion.
• Calculate the cost of a qualified conversation weekly. If it rises — add exclusions, narrow down tasks, refine the chat script.
• Set a 'backward' goal: if 1 in 5 qualified conversations leads to a sale, and the margin is 800 PLN, then a safe cost of a qualified conversation is about 160 PLN (example).

Chat is a conversation, not a list of links. Therefore, in ChatGPT Ads, precision wins: exclusions, specific tasks, and a quick qualification gate in chat. Then, measure only the cost of qualified conversations. If you’d like, I can help you set up these filters and metrics in a short workshop — no coding, using your examples.