I built an AI voice tool for US visa interview practice
Hey IH! I'm building Permito — a platform where visa applicants practice their consular interview by having a real voice conversation with an AI officer.
The problem: millions of people prepare for US visa interviews by reading sample Q&A on blogs. But the real interview is a 2-minute live conversation where the officer reacts to your answers and follows up. Reading doesn't prepare for that.
Permito lets you rehearse that exact experience — pick your visa type (F-1, H1B, B1/B2), enter your background, and talk to an AI officer who asks questions based on your specific situation.
Currently at early stage — a few paying users, growing organic traffic, building out content (blog + free tools like a question bank and DS-160 checker).
Would love to hear feedback from the community. Anyone here been through a US visa interview?
Congrats on the launch, looks solid. How are you currently thinking about acquiring early users and gathering feedback?
Right now my main channels are SEO/blog content and free tools (question bank, DS-160 checker) as lead magnets to drive organic traffic. I also collect feedback directly from users after their practice sessions. Would love to hear if you have any suggestions!
That’s a solid foundation, honestly. Combining SEO with free tools is a smart long-term strategy.
One approach that could complement this is using Reddit to accelerate the distribution of those free tools. Instead of promoting the product directly, you could participate in threads where people are asking questions related to the DS-160 or visa interviews, and naturally reference the tool when it is genuinely helpful.
While this strategy won’t replace SEO, it can provide faster feedback and attract higher-intent users as your organic traffic grows. Are you currently active in any visa or study-abroad-related communities, or are you mostly focusing on search traffic?
This is brilliant! I went through an F-1 interview a year ago and literally practiced in front of a mirror like a weirdo 😅 The pressure of real-time conversation is NO JOKE. Are you planning to add feedback on body language/confidence via video, or keeping it voice-only? Also, the DS-160 checker as a free tool is a great funnel idea
Haha, I know so many people like you — that's literally why I'm building this! 😄 The mirror practice is relatable but it doesn't talk back.
Voice and body language feedback is something the real officer evaluates naturally, so replicating that in the product would be the dream. Definitely going into the backlog — thanks for the suggestion!
Congrats on launching Permito .This nails the real pain: officers interrupt, probe, and you have ~2 minutes to be concise. Do you simulate different “officer styles” (friendly vs strict) and score answers for consistency with the user’s background/DS-160?
Thanks! You nailed it — applicants read sample answers, watch YouTube, feel "ready"... then freeze under real pressure from the officer. That's exactly what we're solving through live voice practice.
Different officer styles (friendly vs tough) wouldn't be hard to implement and we might do it soon, but right now I don't want to overload the tool with features before seeing what users actually need most.
As for consistency scoring — yes, that's exactly what we do. Before the interview, the user fills out a short form with their background, and the AI officer probes for inconsistencies, digs deeper into topics, and sometimes applies pressure — just like the real thing.
Does your system use RAG ?!
Not RAG in the traditional sense — we feed the user's specific background into the conversation context so the AI officer asks relevant, personalized questions. Current models handle this surprisingly well out of the box. But as we scale and refine, RAG with real interview data is definitely something worth exploring!
Congrats on launching Permito 🚀
Visa interviews are super stressful — practicing with real-time AI feedback feels like a huge confidence booster.
Curious, are users getting instant performance insights after each mock session?
Thanks! Yes — within about 15-20 seconds after the session, the user gets a detailed evaluation: what went well, what didn't, where to focus, and how to improve specific answers. We try to make the feedback really actionable, not just a generic score.
That’s great — actionable feedback within 20 seconds is a strong hook.
If you ever visualize those improvements (before vs after answers) it could make the value feel even more tangible.
Really cool direction — feels like something people would happily share after a good session.
Hi there, great product! I lead a portal with approximately 200k social media users focused on immigration to the USA. Are you planning on creating a referral program for your product? I would love to share it on my website.
Thanks, really appreciate it! We're actually building out an affiliate program right now for partners exactly like you. I'll shoot you a DM with the details — would love to get you on the early access list!
Great, looking forward!
yevghenia@gmail.com
This comment was deleted 7 months ago