A lot of reps already use ChatGPT for sales prospecting, whether or not their company has a formal AI tool. That is worth taking seriously rather than dismissing. Used well, it genuinely speeds up parts of the job. Used carelessly, it produces confident-sounding nonsense that can embarrass a rep in front of a prospect. This post is meant to be useful either way, whether you keep using ChatGPT as-is, pair it with something more structured, or move to a purpose-built tool eventually.
The short version: ChatGPT is excellent at reasoning over information you give it, and weak at getting that information on its own. It writes and rewrites well. It does not know your prospect's company unless you tell it, it does not remember your last ten calls with an account unless you paste in the notes, and it will happily fabricate a detail if you do not catch it.
What ChatGPT Genuinely Does Well for Prospecting
Drafting and rewriting outreach. Give ChatGPT a rough idea, a pain point, and a tone, and it will produce a clean first draft of an email or LinkedIn message faster than most reps can write one from scratch. It is also good at rewriting: shortening a message, making it less formal, or adjusting it for a different persona.
Brainstorming angles. When a rep is stuck on how to open a conversation with a specific type of buyer, ChatGPT is a decent brainstorming partner. Ask it for five different angles into a conversation with a VP of Finance versus a VP of Sales, and it will produce genuinely different framings worth picking from.
Summarizing information you paste in. If a rep pastes in a company's About page, a recent press release, or a set of call notes, ChatGPT is good at condensing that into a short summary or pulling out specific facts. This is the use case that actually works reliably, because the model is working with information already in front of it rather than trying to recall or invent it.
Practicing objection handling. Roleplaying a discovery call or an objection-handling scenario is a solid use of ChatGPT, particularly for reps early in ramp who want low-stakes practice before a real call.
Where ChatGPT Genuinely Falls Short
It has no persistent account or lead context. Every new chat starts close to zero. Unless a rep manually pastes in the account history, the deal notes, and the relevant research each time, ChatGPT does not know anything about that specific account beyond what is public and in its training data. That means the rep has to already have done the research and organized it before ChatGPT can help synthesize it.
It requires the rep to know what to ask. ChatGPT is only as useful as the prompt behind it. A rep who does not know which signals matter for a given account, or has not thought through what makes a good opening angle, will not get a good result just by typing "write me a cold email." The tool does not tell you what to look for. It answers what you already knew to ask.
Output quality varies heavily by prompting skill. Two reps using the same tool on the same account can get very different results depending on how well they prompt it. That inconsistency is a real problem for sales leaders trying to standardize quality across a team, because there is no shared workflow, just individual chat histories.
It can fabricate details with total confidence. This is the one that causes real damage. ChatGPT can generate a plausible-sounding but false detail, a funding round that did not happen, a product feature the company does not have, a title that is wrong, and present it with the same confidence as something true. In a sales context, referencing a fabricated detail in an email to a prospect is worse than referencing nothing at all.
It does not scale consistently across a team. Because each rep has their own chat history, their own prompts, and their own habits, there is no shared institutional workflow. A sales leader cannot easily audit or standardize how the team is using it, and onboarding a new rep to "use ChatGPT well" is much harder than onboarding them to a defined process.
It cannot autonomously pull structured account or relationship data. ChatGPT cannot go find a company's recent LinkedIn hires, map out who reports to whom, or pull firmographic and intent data on its own. Anything structured like that has to come from somewhere else and be manually fed in.
A Few Prompts That Actually Work
These are not magic, but they are a reasonable starting point if you are prompting ChatGPT for prospecting work.
Research summary prompt. Paste in a company's About page, most recent press release, and any public info you have, then ask: "Summarize this company in three sentences, then list the three most likely business priorities for their VP of Operations this year, based only on what I've given you."
Angle brainstorm prompt. "I'm reaching out to a [job title] at a [industry] company with [size] employees. Give me five different opening lines for a cold email, each based on a different possible priority this person has, and label which priority each one assumes."
Objection roleplay prompt. "Act as a skeptical VP of Sales who thinks their current process is fine. I'm going to pitch you on [product category]. Push back on me realistically, and after each of my responses, tell me what worked and what didn't."
Notice that all three prompts depend on the rep already having gathered the input, whether that is the company info, the persona details, or the scenario setup. That dependency is the real theme of this post.
The Honest Takeaway
None of this means ChatGPT is bad for sales. It means ChatGPT is a general-purpose reasoning tool, and prospecting is a specific workflow with its own inputs (account research, buying committee mapping, messaging strategy, meeting prep) that the model does not natively have access to. A rep who is good at gathering that context and prompting carefully can get real value out of ChatGPT today. A rep who expects it to already know the account is going to be disappointed, or worse, is going to send something inaccurate.
This is also where Flyfish fits, and it is worth being direct about the relationship rather than pretending they are competitors. Flyfish is not "instead of AI." It is AI that is already structured around the sales workflow, so a rep does not have to rebuild the context and the prompt from scratch every single time. Instead of pasting in company info and hoping the summary is accurate, Flyfish pulls the account research automatically and turns it into the specific assets a rep needs: a messaging angle, a battlecard, a relationship map of the buying committee, meeting prep. The reasoning is similar to what a well-prompted ChatGPT session can produce. The difference is that the workflow, the context, and the output format are already built, so the quality does not depend on how good any individual rep is at prompting.
If you are prospecting with ChatGPT today and it is working, keep using the prompts above. If you are finding that the setup cost, per account, per rep, every single time, is starting to add up across a team, that is usually the signal that it is time for something more structured.
Frequently Asked Questions
Can ChatGPT do sales prospecting on its own? Not fully. ChatGPT for sales prospecting is strong at drafting, summarizing, and brainstorming when a rep provides the input, but it cannot independently search the web for current account signals or maintain memory of a specific deal across sessions.
Is it safe to send ChatGPT-written emails without editing them? No. ChatGPT can generate confident but incorrect details, so any output referencing specific facts about a prospect's company should be checked against a reliable source before it goes out.
What's the biggest limitation of using ChatGPT for prospecting? The lack of persistent context. Every conversation starts fresh, so the rep has to manually gather and paste in the account research each time, which does not scale well across a team or across accounts.
Does Flyfish replace ChatGPT? No, and it is not trying to. Flyfish is a purpose-built workflow that already understands account research, messaging, and meeting prep, so reps do not have to build the context and prompts themselves each time. Some teams use both: ChatGPT for quick drafting and brainstorming, Flyfish for the structured account intelligence and ready-to-use assets.
What should I paste into ChatGPT to get better prospecting results? The more specific and accurate the input, the better the output. Paste in the company's public info, any call notes or CRM data you have, and be specific about the persona and situation, rather than asking a generic question with no context.
