• Examples of disastrous forays with internet advice
  • AI is very convincing – hence dangerous
  • Personal facts remain indispensable

The internet, Google, YouTube, and AI easily (and far too often) provides bad immigration advice – it’s instantaneous, and more literate than a bad attorney, but every bit as risky.

I remember the woman who sat crying in front of me because she’d missed a filing deadline for an appeal. As I handed her back the BIA notice she told me she’d used AI and it gave her incorrect information.

Wanting to save money, a Pakistani national submitted the wrong evidence – incomplete and non-probative – for an adjustment of status application. Before we could refile it, he’d received a “Notice To Appear” (NTA) in immigration court. He was then at the mercy of a weaponized immigration court system.

Another man, a software engineer, was the US husband of an Argentinian woman. He decided to answer a USCIS “Request for Evidence” (RFE) himself, with the help of Google and YouTube. The digital platform misunderstood and misinterpreted the specific eligibility requirements for his wife’s situation; an AI platform even cited an AAO case that didn’t exist. USCIS issued a denial and she was summoned into immigration court. I don’t know what happened to her. Her husband now spends money to defend her in immigration court when he should’ve spent the money to answer the RFE properly in the first place.

What’s the lesson here? – one that Robert Townsend voiced half a century ago: if the input is not precisely correct, automated systems will only “speed up the mess.”[1]

AI Writes Well

When you type a question into an AI platform, you’ll get an answer that sounds convincing, as though it came from an experienced, reliable source. The confident tone is seductive and convincing, isn’t it? When a questionably confident lawyer gives poor advice, you can usually tell by voice inflection or body language that he’s not fully assured that his answer is correct, or even plausible. But though they may too be wrong, AI responses are always assured. Based on none of your specific facts, a dangerously misleading AI answer sounds supremely confident; so confident that you may take it for granted that it’s correct. If you’re looking for a quick fix or trying to avoid legal fees, you may end up in serious trouble. Why? Because Google and AI legal “answers” are incomplete at best, and often categorically wrong. How wrong? I don’t even use AI for even the simplest of concrete facts – such as USCIS fees for petitions and applications.

Case Specific Facts

AI should provide flawless responses for concrete facts – like fees and form edition dates – but they frequently don’t. Client facts – your facts – aren’t absolute. No 2 noncitizen’s have identical case facts. And any accurate legal answer to an immigration question wholly depends on the client case facts. Much can go wrong, viz –

  • A Costa Rican Woman who was perfect candidate for a trafficking visa. AI advised her to apply for a J visa instead;
  • A gay man from Jamaica who had very favorable asylum facts. AI suggested DACA;
  • A US woman and Mexican man recently married, the man looked to adjust status. An AI platform recommended that they provide bank statements, financial data, and estate records. Being married a short time, they had few of them. Before they submitted what they had, I advised them that, since they’d been together for 7 years before they married, there were other ways to easily prove a bona fide marriage. We did that and the man adjusted status.

[1] Townsend, Robert; Up The Organization: How to Stop the Corporation from Stifling People and Strangling Profits, A Knopf publishers, NY, NY, 1970.