The Moment Every Smart Applicant Reaches
You are exactly the kind of person the EB-1A and EB-2 NIW categories were built for. Published research. Patents. A role that mattered inside your organization. You are, by any reasonable measure, good at your job.
So when you sat down to build your self-petition, you did what any capable person would do: you opened an AI model, fed it your CV, and asked it to help you write.
And it worked. The draft that came back sounded professional. It used the right vocabulary — “sustained national acclaim,” “substantial merit and national importance,” “well-positioned to advance.” It felt like real progress after months of feeling stuck.
This is where the confidence sets in. If the AI can produce language that sounds this polished, why would you need to pay a law firm thousands of dollars to do the same thing?
Here is what most people get wrong: the quality of the sentence has nothing to do with the quality of the case. And the gap between those two things is exactly where DIY petitions quietly fail.
Capability Is Not the Same Thing as Control
Immigration attorneys who have watched AI tools evolve over the last two years generally agree on one thing: the capability of these models is genuinely impressive. They can summarize a 40-page research portfolio in seconds. They can draft a personal statement that reads better than what most petitioners could write on their own. They can organize exhibits, suggest headings, and catch grammatical errors a tired applicant might miss at 11 p.m.
But capability is not the same thing as process control — the ability to actively manage a live, evolving legal process and adjust strategy as new information comes in.
A useful way to separate the two:
Capability is what a tool can produce when you give it an instruction. Draft this letter. Summarize this article. Format this table.
Process control is what happens after that draft exists — the judgment calls about what to include, what to leave out, how to sequence the argument, how to respond when USCIS pushes back, and how to adjust the entire strategy when a Request for Evidence (RFE) reveals that an adjudicator is reading the case differently than expected.
A DIY applicant using AI has capability. What they are missing — and what they often do not realize they are missing until the RFE arrives — is control over the process itself.
The Common Mistake: Treating a Draft Like a Strategy
Here is the pattern we see most often at Silmi Law when a DIY petitioner comes to us after receiving an RFE or a denial. It is rarely that the writing was bad. Often, the writing was quite good.
The problem is almost always structural, not stylistic. The applicant built a document when what they needed was a legal argument.
An AI model, no matter how advanced, does not know:
- Which of your ten possible pieces of evidence actually satisfies a specific regulatory criterion under 8 CFR 204.5(h)(3), and which ones are decorative
- Whether your proposed endeavor, as framed, actually threads the needle required by the three-prong Matter of Dhanasar (2016) test for NIW
- Whether an adjudicator reviewing your service center this quarter is scrutinizing “original contributions” harder than “critical role,” based on nothing you can see from outside a law practice
- Whether the letter from your former manager actually establishes a “leading or critical role,” or simply repeats your job title in different words
This is the trap: a fluent draft feels like a finished case. It is not. It is raw material. What turns raw material into an approvable petition is judgment — about sequencing, about emphasis, about which three or four criteria to build a case around instead of scattering thin evidence across six.
Where AI Genuinely Helps — and Where the Wall Appears
To be fair to the tools: AI is not the enemy of a strong self-petition, and no responsible attorney pretends otherwise. Used correctly, AI is useful for a narrow set of tasks.
Where it helps:
- Organizing your accomplishments into a rough timeline before your first attorney consultation
- Drafting a first-pass summary of a technical publication so a non-technical reader can follow it
- Catching typos, inconsistent dates, and formatting issues across a long exhibit list
- Helping you remember details you might otherwise leave out of a first draft
Where the wall appears — every time:
- It has no visibility into how adjudicators are behaving right now. An attorney who filed a petition two weeks ago and received an RFE knows, in real time, what a specific service center is pushing back on this quarter. AI models are trained on data with a cutoff date, and immigration policy, service center behavior, and adjudication trends shift constantly — sometimes month to month.
- It cannot see the pattern across your specific case. A strong attorney reviewing your background will ask a follow-up question about a conference you mentioned in passing, connect it to a publication you forgot to list, and realize that connection is the strongest evidence in your entire case. AI answers the question you asked. It does not know which question you should have asked.
- It has no accountability for the outcome. If an AI-drafted argument is wrong, nothing happens to the AI. If your petition is denied because a legal standard was misapplied, you lose your priority date, your filing fee, and months or years of your immigration timeline.
- It tends toward hedged, generic language. Persuasive legal writing takes a position and defends it. Many AI-generated drafts, especially from models tuned for caution, soften claims instead of committing to them — which reads to an adjudicator as an unconvincing case, not a safe one.
None of this means AI is useless. It means AI has capability without control, and self-petitions are won or lost almost entirely in the control layer.
Why This Matters More for EB-1A Than People Expect
The EB-1A extraordinary ability category is deceptively structured. USCIS uses a two-step analysis known as the Kazarian framework: first, does the petitioner meet at least three of the ten regulatory criteria; second, considering the totality of the evidence, does the petitioner belong to that small group at the top of their field.
Most DIY petitions fail at step two, not step one. It is not hard, with AI assistance, to generate documentation that technically checks three boxes — a few citations here, an awards certificate there, a judging letter. What is hard is building a coherent, cumulative narrative that convinces an officer you are genuinely among the best in your field, not simply someone who found three ways to check a regulatory box.
That narrative requires knowing, case by case, how much weight an adjudicator is likely to give a particular type of evidence today, how to frame a “critical role” so it survives scrutiny, and which criteria are worth building deeply versus which are worth mentioning briefly. That is strategy. It is not something a prompt produces.
Why This Matters for the EB-2 NIW
The National Interest Waiver runs on the three-prong Dhanasar test: the proposed endeavor has substantial merit and national importance; the petitioner is well-positioned to advance it; and, on balance, it would benefit the United States to waive the labor certification requirement.
The most common DIY error we see is prong one: applicants describe their field as important (“AI is important,” “renewable energy is important”) instead of their specific endeavor within that field. USCIS has been increasingly precise about this distinction, and an AI model asked to “explain why my work is nationally important” will often produce exactly the generalized answer that gets flagged in an RFE — because it is answering the question literally, not answering the question an adjudicator is actually asking.
Capability vs. Process Control: A Side-by-Side View
| AI Tools (DIY) | Attorney-Managed Strategy | |
| Drafting speed | Fast, available instantly | Slower, deliberate |
| Knowledge of legal standards | General, based on training data | Current, based on active case practice |
| Visibility into adjudicator trends | None | Direct, from RFEs and approvals in real time |
| Ability to ask a clarifying follow-up | Limited to your prompt | Draws out details you didn’t know mattered |
| Adjusts strategy after an RFE | Cannot — no memory of your live case | Rewrites argument based on what officer flagged |
| Accountability for the outcome | None | Professional responsibility for case quality |
| Evidence weighting judgment | Treats all evidence as equally useful | Prioritizes criteria most likely to persuade |
| Cost if the case is denied | Lost priority date, filing fee, and time | Mitigated through strategic filing from the start |
Self-Assessment: Are You Actually Equipped to DIY This?
Answer honestly before you file.
☐ I can explain, without checking anything, exactly which three (or more) EB-1A criteria my case satisfies and why each piece of evidence meets the regulatory definition — not just the plain-English one
☐ I know how current USCIS adjudication trends are treating my specific field this quarter
☐ I have reviewed at least one real RFE or denial in my field and understand what it reveals about how officers are reading cases like mine
☐ My “proposed endeavor” (for NIW) names a specific initiative, not a general field of work
☐ I have a clear plan for exactly how I would respond if I received an RFE — not just intent to “figure it out then”
☐ Someone other than an AI model has reviewed my case and pointed out a gap I hadn’t considered
If you checked fewer than four boxes, your case has capability behind it but no process control. That is precisely the gap where strong candidates receive avoidable RFEs.
Case Study: When a Strong Profile Still Needed a Strategist
Details altered to protect confidentiality.
A senior data scientist at a mid-size technology company came to Silmi Law after receiving an RFE on a self-filed EB-2 NIW petition. The petitioner had used an AI model extensively to draft the proposed endeavor statement, and on its face, the writing was fluent and well-organized.
The RFE challenged prong one: USCIS found the proposed endeavor “not sufficiently specific” and questioned whether the described work rose to the level of national importance.
On review, the underlying case was strong. The petitioner had led development of a fraud-detection model adopted across multiple financial institutions, with measurable impact on transaction losses. But the AI-drafted statement had described the endeavor as “improving machine learning for fraud prevention in the financial industry” — a framing broad enough to describe hundreds of engineers’ jobs, and exactly the kind of generalized language that draws RFEs under current Dhanasar scrutiny.
Silmi Law restructured the endeavor around the specific model, its measurable adoption, and its quantifiable impact on financial institutions’ loss rates — reframing a generic field description into a concrete, evidenced initiative. The RFE response was filed within the response window, and the case was approved.
The lesson: the underlying qualifications were never the problem. The AI-generated draft had capability — clean prose, correct vocabulary — but no process control over how that specific USCIS office was reading prong one that quarter. That distinction was the difference between a denial and an approval.
Frequently Asked Questions
Can I use ChatGPT, Claude, or another AI tool at all while working with Silmi Law?
Yes. We regularly review client-drafted material, including AI-assisted drafts, as a starting point. The issue is never the tool — it’s relying on it as the final strategist rather than a first-pass assistant.
Is DIY filing ever appropriate for EB-1A or NIW?
USCIS does not require attorney representation, and some straightforward cases with very strong, well-documented records can succeed without one. The risk is not that DIY filing is impossible — it’s that most applicants cannot accurately assess, from the outside, whether their case is actually one of those straightforward ones.
What happens if my AI-drafted petition already went out and I got an RFE?
An RFE is not a denial, and many strong cases are saved at this stage. What matters is responding with a legal argument tailored to exactly what the officer flagged, not a general clarification of the original draft.
How is Silmi Law’s process different from just using better AI prompts?
Prompting produces better drafts. It does not produce visibility into current adjudication trends, does not ask you the follow-up question that surfaces your strongest evidence, and does not carry professional responsibility for the outcome of your case.
Does using AI to draft my petition hurt my case with USCIS?
Not inherently — USCIS does not ask how a petition was drafted. What hurts a case is submitting generalized, unpersuasive, or legally misaligned language regardless of its source, whether that language came from an AI model or from an applicant writing alone.
What This Means for You
Being smart enough to use AI well is not the same thing as being equipped to run your own legal strategy against a system that changes its posture every quarter, reads evidence against standards you don’t see from the outside, and gives you exactly one real chance to make a first impression with your priority date on the line.
At Silmi Law, we work with engineers, researchers, physicians, and founders across AI, biotechnology, semiconductors, energy, and the arts who are entirely capable of drafting their own material — and who still come to us because writing a good sentence and building an approvable case are two different skills. We don’t discourage clients from using AI tools to prepare. We make sure what gets filed reflects real-time knowledge of how these cases are actually being adjudicated, not just how they read on the page.
If you’re weighing whether to file on your own, the question worth asking isn’t whether you’re capable of writing a strong draft. It’s whether you have control over everything that happens after you hit submit.
Schedule a Strategic Case Review
An informed decision made before you file is worth far more than a corrective strategy made after an RFE or denial.
- Schedule a Free Consultation — 15 minutes, conference call
- Schedule a Consultation with Attorney Sharif Silmi — 30 minutes, $295, video call
- Schedule a Consultation with Attorney Sharif Silmi — 60 minutes, $495, video call
- Schedule a Working Session with Attorney Sharif Silmi — 75 minutes, $895, video call
Legal Disclaimer
This blog post is provided for informational purposes only and does not constitute legal advice. Reading this content does not create an attorney-client relationship with Silmi Law Firm.
Immigration law is highly fact-specific and complex, and USCIS adjudication standards and trends referenced in this article are current as of the publication date but are subject to change without notice. The information here represents general principles and should not be relied upon as advice for your specific circumstances.
Before taking any action related to your immigration status, EB-1A, or EB-2 NIW petition, you should consult with a qualified immigration attorney, provide complete information about your individual circumstances, and obtain written legal advice tailored to your case. Silmi Law makes no guarantee of approval and no representation as to the outcome of any specific case.
For legal representation, contact Silmi Law: Phone +1 (443) 329-2929 · Email info@silmilaw.com
