ChatGPT is my new college advisor and it’s so much easier than working with a person

[Anonymous]

So I’ve been trying to find a college advisor for my sophomore DS and I’ve had a couple of clunky meetings with a couple people from a large company. It then fell through for a number of reasons and I’ve decided to ask charGPT about a fairly uncommon field (in humanities instead of stem that everyone in my area is crazy about).
Well it did give me a few school options with FAFSA as opposed to CSS and relatively affordable COA. of course it all needs to be double checked but at least I have maybe 5 schools to look at with my kid. Surprisingly they aren’t even that competitive which is great in our case as he is not a stellar student, just average (GPA around 3.5 and only 1-2 honors/AP classes per year).

Anyway, I realize the limitations but I just wanted to put it out there that AI can be very calming for someone like me, a parent who knows nothing about their child’s preferred field and is anxious over all the choices and their kid’s academic achievement (or lack thereof). Sorry if I’ve reinvented the wheel but just wanted to share!

[Anonymous]

I don't think you need to use a professional college advisor to have great college results.

However, I would remain skeptical of whatever chatGPT is spitting out. One of my kids used chatGPT to get an initial list of grad programs and there was an absolute ton of incorrect information coming out of chatGPT. It's just so unreliable.

[Anonymous]

You need to implement a prompt stopping chatgpt from hallucinating every time you ask a question.


Search up on exemplary stop hallucinating prompts on google. It works.

[Anonymous]

Anonymous wrote:I don't think you need to use a professional college advisor to have great college results.

However, I would remain skeptical of whatever chatGPT is spitting out. One of my kids used chatGPT to get an initial list of grad programs and there was an absolute ton of incorrect information coming out of chatGPT. It's just so unreliable.

True, I realize I need to check each program individually but at least there’s a list to work off of.

[Anonymous]

Anonymous wrote:You need to implement a prompt stopping chatgpt from hallucinating every time you ask a question.


Search up on exemplary stop hallucinating prompts on google. It works.

Thanks, I will research! I’ve already noticed that he first gives me an example of a program and then in the next iteration tells me it’s closed as of mid 2025. I am sure there are more snafus like this but it’s a great start and very calming

[Anonymous]

Right. It’s generative — so it’s really really good at making something from nothing — and really helps when the empty page is the problem.

It also tells us what we want to hear and makes shit up — so it’s an ok starting point, but then be sure to step away

[Anonymous]

As long as you know there are possible hallucinations, I think it can be a great starter point. But it has trouble really putting together context and tends to rule by consensus, and therefore never seems to give 'exception to the rule' analysis.

So you just have to double check its work -- think about the lawyers who have been disbarred for submitting ChatGPT legal briefs with made up cases in it.

[Anonymous]

Anonymous wrote:Right. It’s generative — so it’s really really good at making something from nothing — and really helps when the empty page is the problem.

It also tells us what we want to hear and makes shit up — so it’s an ok starting point, but then be sure to step away

Ive been using GPT-5 paid version and its better than it use to be. Set up a project with instructions to be a critical/jaded AO that speaks with brutal honesty. It's trained on a ton of information from reddit, forums, interviews, school websites, etc. So while it's a 'word completion' model its predictions are based on its training. Plus you can ingest data and play with that as well.

But, yes, one has to remember its not 'thinking'. It's a model. It does get hung up on certain outputs or regurgitate stereotypes unless challenged but its gets better output after refining prompts. So to your point, it can enhance ones research, but don't trust it blindly.

[Anonymous]

Yes, remember that ChatGPT or any other generative AI is not multifactor Google that narrates, as much as people wish it were and try to use it that way. It can talk and talk and talk without having complete information or even good information. Check the AI summaries you now get from Google against the actual search results: even that will show you how bad it is. I am a career researcher and I can't trust consumer-facing AI for anything at all that I've ever tried.

[Anonymous]

Considering how much it has been trained over the last 3 years with applicants and their parents using it for college admissions, it is likely it is very good by now.

[Anonymous]

Anonymous wrote:Right. It’s generative — so it’s really really good at making something from nothing — and really helps when the empty page is the problem.

It also tells us what we want to hear and makes shit up — so it’s an ok starting point, but then be sure to step away

Exactly, it helped me with the empty page problem.
The two people from the big company gave us random school names, most out of state, for summer programs to look at. Neither of them said hey there are 3 nice schools near you that have the program that is a potential fit for your kid, and since you told us you prefer publics due to FAFSA reasons, they are the ones you should look at.
They aren’t too competitive either as your kid has an average GPA as of now.

[Anonymous]

I think that's a great idea, OP.

[Anonymous]

When I use goggle, the AI generated answers are often incorrect or contain errors. Not sure if this is ChatGPT or another AI program.

[Anonymous]

Anonymous wrote:When I use goggle, the AI generated answers are often incorrect or contain errors. Not sure if this is ChatGPT or another AI program.

Every year the models are retrained. Some are focused on one thing over another.

[Anonymous]

Thanks, OP. Great advice.