Self-study
We just rolled out our AI Coach to every user on 7Sage. Use it to analyze your practice results and get personalized recommendations on how you can improve. Soon, it will also be able to explain tricky LSAT questions (using 7Sage explanations as the source—we're not allowed to show it LSAT content.)
You can access it by opening the chat menu or clicking any of the AI buttons in Analytics:
Please leave feedback in the comments!
28
68 comments
@David_Busis A quick comment on an error it keeps producing - I do like it and find it valuable, BUT - it keeps mis-reading/doesn't understand when you have adjusted your timing. I have accommodations for testing for 150% time, so I typically practice with those conditions. But then if I ask it for analysis, typically it will bring up that I'm practicing in "standard" timing and I'll do better under my expected timing, which is just not true. It will say this when I've done a section on 150% time.
Also, agree it's sycophantic like all AI models - which is nice when I need a little confidence boost to keep going when I get something wrong! But too much, and I worry it can actually derail people's progress if they have some false hope/think they're doing better than they actually are. Which could end up being worse if you start slightly prompting it to boost yourself with the questions you ask. I noticed this when I asked after an LR section, "How would this influence my overall score because I'm already very strong in RC?" And it instantly agreed with that, even though - yes I've gotten perfect passages and sections where I only get -1 or -2 in RC, but have also had some ones with as much as -6. And I would hate for it to keep giving me the impression that "it's ok - you got this still!" when I actually have some areas I really still need to improve.
My sister tutors for the SAT and says a really common mental trap for students is to make excuses for themselves when they didn't get something right they already feel confident in ( "I got distracted briefly on that question..," and "oh I just read that one too fast...", "I have a headache today, that's why it was wrong.." etc. )- those things could very well be true, but I could see the AI reinforcing some of these thoughts based on prompts you put in, possibly to your detriment.
I'm a critic of AI, but I do think it has some very limited useful applications and see how this could be improved to be very useful to students to continue to identify learning gaps and help with study strategy, but I would be very careful about monitoring and fine tuning the product to trim where it could actually be hurtful to students.
@EEC_517 Re: accommodations we're aware of the issue and fixing it.
@J.Y.Ping Great thanks!
To me, LSAT studying has two goals: 1. achieving a desired score, and 2. sharpening my reasoning skills. I subscribe to 7sage because the content is high quality and is more likely to serve my goals than the alternatives. I realize the drive to release an AI product is very difficult to resist, especially for a company that sees creating new products as one of its fundamental roles, but I am still disappointed to see it. In my experience, LLMs universally degrade quality relative to human work product. Empirically, they can produce higher preference scores in some contexts, but that is not quality. It's related to sycophancy. I think this will be exciting for a number of people and might even lead to more customer satisfaction, but I do not expect it to produce a measurable benefit for the primary goals of LSAT studying, and it may produce a harm. I expect the sort of metrics you will use to evaluate this product are engagement and satisfaction, in addition to net cost. Please think carefully about how to test the effect of this tool on the quality of the content you produce and the effectiveness of your product on student achievement as well.
@dh2303 You can turn it off at the bottom of your settings: https://7sage.com/users/settings/preferences
@J.Y.Ping respectfully, that is not really responsive to my concern. My concern is not that seeing your AI product is going to harm me specifically, but that without an atypical measurement strategy, your AI product is going to harm 7sage. Management attention is finite. AI products are likely to generate compelling business metrics that are easy to measure (engagement and satisfaction), and, in my opinion, likely to produce harms on the 7sage system in general, which are harder to measure (reduced teaching effectiveness and reduced objective quality in total content, when compared to the counterfactual without an AI product). It's your company, of course you can do what you want, but I think your commitment to quality created a uniquely powerful education product. I've made a specific suggestion about how to protect that despite the pull of AI in edTech.
It is extremely rare that I find an AI feature that actually helps me, however, in the limited scope of (1) spitting raw data out to me in a plain-English format, and (2) providing study strategies based on my performance and focus thus far, I like the implementation of the AI Coach.
I really enjoy the way it analyzes specific LSAT questions, because it DOESN'T. The AI is prohibited by LSAC from breaking down the actual questions (makes sense for many reasons), which mitigates some risk of inadvertent over-reliance on AI or using it to think for me. Honestly, if the AI Coach actually broke down questions or summarized 7Sage explanations, I wouldn't use it at all.
My biggest critiques are, unfortunately, inherent to all LLMs - overly wordy and tends to unnecessarily repeat itself, sycophantic, and anything and everything it says must be double-checked by a human.
Overall, it's good when used purely to crunch data and get recommendations on best study practices. Otherwise, no thank you, I would rather do it myself.
Detailed response examples:
Again, I very much appreciate that the AI Coach is prohibited from analyzing the questions themselves. As an experiment, I entered the prompt: "Analyze my incorrect science RC answers since March 1, 2026." The AI Coach responded with:
It followed this disclaimer with a couple of tips on best practices for reviewing and analyzing these questions myself:
I have also found its analysis of specific sections/drills/PTs helpful. Because it cannot analyze the questions themselves, I am only getting info available to me through the Analytics page, just in an easily digestible format which points me toward strategies and types of drills that I may want to focus on. Which is great! I don't want the AI to think for me - I only want it to point me in the right direction. For example:
Lastly, I like the suggestions it gives when you give it the prompt "What should I do next?" For example:
[This comment was deleted.]
@maydhak I think at the very least, it does not automatically trigger like it does with Google searches or things like Spotify or Meta. You have to talk to it for it to work.
Please make the lesson plans more customizable. Like able to block out certain dates and it being able to adapt if you miss a day!! My OCD cant take this lmfao
[This comment was deleted.]
@ionicinstinct Bravo
@ionicinstinct To be honest, I am unsure if I agree that AI has relevant limitations. I have Gemini Pro from class, and it has never gotten an LSAT question wrong or given me incorrect logical reasoning. Obviously that doesn't guarantee anything (this is not a flaw stimulus, save it for the test this weekend) but I have found it very helpful for when a 7Sage explanation feels too post-hoc for my liking.
@antitrust_fan I figured that other models would fare differently. I also found out since that the AI coach does not allow specific LSAT questions, so it's kind of irrelevant. But given anecdotal instances of Geminis hallucinations and articles about Gemini's hallucations, I would still say the point stands in regard to the rest of the AI coach's tasks, just a little more weakly: That there's curerntly limitations to AI that put it at risk of innacurate answers. Either way,
@antitrust_fan So you're using it to review missed questions after you've read/watched the 7Sage explanations and lauding it for never being wrong? Please be a bit more honest with yourself.
@icedshakenespresso
You first.
Not only have I used it to clarify or expand on 7Sage explanations, but I have also given it questions where it has not seen the correct answer or 7Sage explanation and have watched it give me a more refined version of the 7Sage explanation.
The cynic in you could very well argue that maybe some Google engineer stole 7Sage's explanations and uploaded them into the model, but neither of us has any evidence of that. It also probably wouldn't be nearly as good at mapping or coming up with its own unique examples (with identical logic) if it was just regurgitating things from the internet or hallucinating.
I am critical of AI, but I recognize there are some things it can do and some things it cannot do. It cannot write a cover letter. It can trace conditional logic.
@antitrust_fan Stop. All LLMs hallucinate. Your anecdotal experience of plugging in a couple of questions is not evidence of anything.
Huh? Gemini doesn't need to be trained on 7Sage materials; it's trained on a colossal amount of data. The internet is full of discussions of LSAT PrepTests, Reddit in particular. Do you understand how this works?
@icedshakenespresso
Please point out where in my statement I said, "Gemini does not hallucinate."
And because you have such a strong distaste for anecdotal evidence, here are AI models getting a perfect score or near-perfect score without access to the materials: (Ku 2026)
Do you understand how this works? It can reason.
You are jumping on the AI hate-train for one of the few things it can actually do.
And before you give another rhetorically charged yet virtually substance-free reply, ultimately all that needs to be true is that the AI is a little better than Reddit/7Sage comment sections; otherwise, you should be advocating for getting rid of the comment sections too due to their unlabeled inaccuracy. God forbid a student pick up a bad habit from another student's comment on an LR question.
@antitrust_fan LLMs do not reason. This is not up for debate. Their ability to recognize statistical patterns in syntax is limited to what they've been trained on. How exactly do you think these engines improve?
Yes, we all know AI has scored a 180. The LSAT is a multiple choice exam. This does not support your claim that they are able to reason. I'm guessing you believe calculators reason as well?
There is no "AI hate train". Being a principled person is not "hating". Your LLM is giving you Trumpian Twitter arguments.
Literally what are you on about? Comment sections are a place for discussion among peers, not a place where students expect to find flawless explanations. "Rhetorically charged yet virtually substance-free"? Using pithy Redditspeak doesn't give you any intellectual high ground.
@icedshakenespresso
Not too sure I feel the Reddit insult lands from the one whining about AI all over this thread “dunking” on anyone who disagrees with them with what I can only describe as mean-spirited replies that are “rhetorically charged” talking points everyone has already heard.
I’m sorry, but I just don’t think you know more than Harvard CS professors. I am astounded that you think you do. Please do not go to the same law school as me.
@antitrust_fan Right, the only person I "dunked on" was someone using ChatGPT to critique the "passive aggressive" language of 7Sage's AI chatbot. This is crippling addiction no matter how you slice it and it should be shamed. Don't forget that you and I are only having this conversation because you responded antagonistically to someone sharing their perspective.
The people profiting from AI usage and investment want you to believe it can "reason". This is literally just a semantic point, it depends on how you define "reason". Nothing I said was untrue. If you had any idea how AI worked this wouldn't be so difficult for you. Please continue defending the likes of Palantir, I think it'll be really fulfilling for you.
Yes, I'm clearly the mean-spirited one here. If you want to continue to use AI to get through life, I'm certain we'll end up at different schools. :)
@antitrust_fan You never answered my question. Can calculators reason?
@antitrust_fan I don't want to get in a flame war, but you're drawing an unwarranted conclusion about what LLM AIs can do well. They do well on benchmarks that test on exam material that is present in their training set (in other words, they are good at knowing the answer if they already have that answer in their training set). This is not a useful measure of being able to produce output that properly analyzes, e.g., the conditional logic. It's polluted with the answer choices and it's polluted with the large volume of low quality analysis
@icedshakenespresso No LLM has "scored a 180 on the LSAT." It has scored a 180 on an exam that includes question sets present in its training data, by virtue of being on the open internet. LSAC has not offered any AI company a live look at question sets not published.
@icedshakenespresso @antitrust_fan I get email notifications for every post I don't know how to turn off. I already deleted the comment. Could we please put this to rest?
@dh2303 Okay, thank you clarifying and for backing me up.
I probably won't utilize this much but I tried playing around with it and it highlighted that I skipped blind review on questions that the blind review didn't suggest I go over again. That would be a reasonable note but I've skipped blind review at most a few times besides when I'm doing the adaptive drill function that doesn't even have blind review.
I think it was even more pointed as it complained about not doing blind review on a recent drill despite me doing a blind review on the suggested questions. I suppose you could say I could blind review every question but if I have it set up to suggest questions for me, there's clearly significantly less benefit of me trying to "blind review" a question I'm being told I got right by nature of it not being flagged for blind review.
@Adam Facing this too. When doing the adaptive drills, I’m immediately told whether I got the answer right or wrong and what the correct answer is (is there a toggle for this?). The AI every time tells me that I should be doing blind review on these problems, but how can I do blind review when it tells me the answer? I asked, and it said:
“Good question — and it's a fair frustration with how the adaptive drills flow. The key is to do your "blind review" thinking before you submit each answer, since you're right that the correct answer appears immediately after.
Build the review into your first pass. On any question where you're not 100% confident, pause before submitting and force yourself to articulate:
Why the answer you're choosing is right
Why each of the other four is wrong
If you can't cleanly do that, flag it mentally (or note the question number) as one to revisit even after you've seen the answer.”
Not my favorite response, tbh.
@Adam @Charlie03 thank you! We just pushed a fix that should address these spurious "do more BR" replies.
I am really loving it! However, just a few thoughts:
1.Yeah it’s a bit…sassy. Personally, I like it because it pushed me to do the blind reviews but will other people like it, not too sure. Maybe add something that allows us to change the tone from nurturing to aggressive that way we can pick what we like. Some of us need that extra push.
2.It would be cool if it had links sending me to the section it thinks I need to work on. For example:
“So your two targets from this drill are — casual weakening and necessary assumptions — are exactly the LR areas already on top of your priority list”
Once we click on the bolted parts, we could then be linked to the exact lesson we need to go back to review.
3.Visually it’s too long. Perhaps it can separate what it’s giving us into sections. First paragraph criticize my timing, second paragraph discuss my second guessing, third paragraph discuss my blind reviews and fourth paragraph talk about where I need to focus and provide me the links to the lessons. Give it spacing and maybe bolded titles so we can be able to distinguish the type of feedbacks we’re getting.
Ultimately, it’s awesome! I know it’s in beta right now and I hope it can get to a point where it helps us create study schedules, and study goals for the week and such since it has access across the platform I imagine.
@Crisgee12 Thank you! We're working on links. We just pushed a fix that should make the responses more skimmable.
It keeps giving giant paragraphs that are walls of text instead of concise bullet point breakdown feedbacks even after I explicitly told it to…Would also like option to clear chat history or start new chat.
@RinHan On it!
Super super useful
I really love it, but sometimes it has passive aggressive vibes: "Lidiia, you skipped Blind Review again. This is the third drill in a row where you have done this, and at this point it is the single biggest thing holding back your progress. Please go back and Blind Review Questions 4 and 5 before moving on." chatgpt answered that: The phrases that create that feeling are:
"you skipped Blind Review again" — "again" can sound exasperated or scolding.
"This is the third drill in a row" — keeping count of failures can feel like being reprimanded.
"at this point it is the single biggest thing holding back your progress" — a strong judgment that may come across as blame rather than observation.
"Please go back ... before moving on" — an instruction immediately after criticism can feel parental or disciplinary.
@Lidiia You are under spells
@icedshakenespresso leave it
@Lidiia You're complaining about a chatbot having "passive aggressive vibes"...
@icedshakenespresso off topic
@Lidiia You critiqued the AI . . . by using AI to write the criticism?
@Lidiia Ratioed btw
@DavidDuncan88 These are future legal professionals, can you believe it?
@DavidDuncan88 This AI is trained by 7sage materials, they were specifically saying that during their podcast, so the way this Chat replying it was taught by human and it need to be corrected more neutral or positive tone.
@icedshakenespresso text can have attitude. See, e.g., RC questions, which ask you to infer attitude from text. It is entirely appropriate to share, in a forum about a new product, the attitudinal qualities of text output from that product.
actually a lot more insightful than i thought! i thought it was going to give me generic study plans but it takes quite a few factors into account
This is... an odd choice.
@icedshakenespresso You can turn it off at the bottom of your settings: https://7sage.com/users/settings/preferences
It can even build you LR and RC drills!
Hello @David_Busis,
I am waiting for your team to roll out a 7sage offline feature. It would be very useful to study on planes, trains, and other forms of transportation where connection is not always stable. If I can pay $15/month for Spotify and have access to all locally downloaded content within the app without an internet connection, I am wondering why this $75/month service cannot perform similarly. Fingers crossed.
#feedback
@AOnifade I would also love to see this. I commute to school a lot, and around the northeast US in general, and I constantly think about how useful that time would be if I could study LSAT during it.
I'm currently playing around with this, and it's pretty cool! Are you considering granting the AI access to comments we make on questions? I'd be curious to see what insights it could pull from my wrong answer journal comments.
I have genuinely been waiting to see if you guys implement this feature. Haven't tried it yet but my ultimate hopes as a user would be if it could tell me more about specific things like whether I misread double negatives and pedantic stuff like that as I do thousands of questions.
@antitrust_fan You undersold this! Admittedly, I didn't watch the video, but it's so cool that it can build drills. The only thing I noticed about the analyze feature was that it weirdly only will analyze what's on the screen. (It told me that there were only tags on 7Sage regarding passage type for RC, and there was actually data for certain RC questions, but they just happened to not be on the default analytics page).
While I hope to be done with 7Sage for good this June, I am pretty impressed with this tool, especially for someone who is often skeptical about implementing these things.
However, you should probably know that I asked it to make a drill with a specific passage (I gave it PT105.S3.P2.Q1 in the proper formatting), and it gave me something pretty random with full confidence.
This is such a game changer. It's essentially a tutor built in to 7sage. I love it. And I'm already seeing improvement from it.
❤️ we already use it outside of 7Sage just with caution as it’s not fully vetted or customized to the lsat as a whole…. Intrigued to see if there are flaws with the full integration. Outside of 7 sage on multiple occasions ai could not process high level logic…
Pending to see as of today we embrace it⭐️
I am very excited to cancel my 7Sage membership. AI is terrible for the environment and questionable at best for humanity.
@charlenehansen and AI is terrible at legal reasoning, which is why law schools like UC Berkley are banning its use. Idk why you'd wanna go through so many hoops (prompt upon prompt) for AI to get a perfect 180. We know it's possible, but I imagine we would not hire a tutor that scores anywhere between 149 (GPT-3.5) to 163 (GPT-4). I, for one, enjoy humans who can explain why they got the human-ly scored 170+
all that to say, @everyone please get help from humans. Hallucinated logic is flawed logic
@cattail8148 I love you
@icedshakenespresso <3 u are my favorite dunkin drink
@charlenehansen I wouldn't go so far as to reccomend deleting 7sage, as my points have gone from 150s to 170s/high 160s, and its very useful. But definitely try to reccommend against usage more constructively. I'm surprised people aren't pointing out the biggest issue, which is cognitive offloading leading to worse critical thinking - dangerous for people taking the LSAT. @icedshakenespresso @cattail8148 Thank you for being the only others in this thread recognizing how harmful this is, because I was beginning to lose hope in people looking at the comments.
@ionicinstinct im almost finished with my lsat journey so it is no biggie for me
@charlenehansen good luck!
@charlenehansen You can turn it off at the bottom of your settings: https://7sage.com/users/settings/preferences
It's amazing. I wish I had it before my test though :(
Great so far! I wish I could open different instances of the chat so that I don't have to scroll up and down a long conversation to find a particular response.
AGH THIS IS WHAT I WISH I HAD 2 MONTHS AGO
Love it so far!