You’ve probably run your numbers through two or three screeners by now, gotten two or three different answers, and come away with the nagging suspicion that none of them actually know what they’re talking about. That suspicion is correct, and it’s worth understanding precisely why—because the gap between what a screener produces and what a caseworker produces isn’t a bug or an inconsistency between websites. It’s a structural difference in what the two processes are designed to do.
What Counts as a Legal Determination
A legal determination is a specific, bounded event. It happens when an agency—a state SNAP office, a Medicaid enrollment unit, a Marketplace exchange, a utility assistance program administrator—reviews a submitted application along with supporting documentation, applies the current rules for that program and that jurisdiction, and issues a decision. That decision has a few characteristics a screener result never has:
- It’s tied to a specific household on a specific date, using the rules in effect on that date.
- It’s based on verified information: pay stubs, tax documents, proof of residency, household composition as documented rather than self-reported.
- It comes with procedural rights attached—the ability to appeal, request a fair hearing, or submit additional evidence if something was miscounted.
- It’s recorded. The agency has a record that this determination was made, and that record has legal weight if a dispute arises later.
None of that exists on the other side of a screener. When you type your household size and an approximate income range into an online tool and it tells you that you’re “likely eligible” or gives you an estimated benefit amount, nothing has been reviewed, nothing has been verified, and no agency has taken any action. You have not applied for anything. No one downstream even knows you used the tool. The screener has simply run your inputs through a set of rules and returned a probability-weighted guess.
This distinction matters more than it sounds like it should, because the language screeners use is often borrowed from the language of actual determinations—words like “eligible,” “qualify,” and “approved” show up in both contexts, and most tools don’t work hard to signal which meaning they intend.
Why Screeners Are Built to Approximate, Not Confirm
It helps to think about what a screener is actually trying to solve. Its job is to take a small number of quick inputs—household size, rough income, maybe age or state of residence—and produce a fast, low-friction estimate for someone who hasn’t yet decided whether applying is worth their time. That’s a genuinely useful function. Full applications are long, and a tool that helps someone decide “yes, this is worth twenty minutes of my time” or “no, I’m well outside the range” saves real effort on both sides.
But that function requires the screener to compress a lot of complexity into a handful of fields. Real eligibility rules for programs like SNAP, Medicaid, or ACA subsidies involve variables that a five-field form can’t capture cleanly:
- Deductions and disregards that depend on specific expense categories, not just gross income.
- Household composition rules that differ from how people naturally describe their living situation (who counts as part of the “household” for program purposes is not always who shares an address).
- Asset tests that some programs apply and others don’t, and that vary by state.
- Timing issues—income averaged over a period, versus income as of the date of application.
- State-specific variations layered on top of federal baseline rules, especially for Medicaid and LIHEAP-type programs.
A screener that tried to model all of this precisely would stop being a screener—it would become an application, with all the friction and time investment that implies. So every tool makes a design tradeoff: ask fewer questions, accept less precision, and communicate a result that’s meant to be directional rather than definitive. The people building these tools generally know this. The problem is that the tradeoff isn’t always visible to the person using the result.
There’s also a quieter reason screeners lean toward approximation: the underlying rules change. Income thresholds get updated, program parameters shift, and a screener maintained by a third party may be running on a rule set that’s slightly out of sync with what the actual agency is using this month. A tool that’s ahead of or behind a rule update by even a short window will produce a confident-looking answer that’s confidently wrong—not because anyone did anything careless, but because eligibility rules aren’t static, and keeping every screener perfectly synchronized with every program’s current parameters is harder than it looks from the outside.
How Results-Page Language Signals Which Kind of Answer You’re Looking At
Since most screeners don’t explicitly announce “this is an estimate, not a decision” in large type, you have to read the phrasing on the results page itself to figure out what you’re looking at. There’s usually a tell, if you know where to look.
Words that suggest you’re looking at an estimate, not a determination:
- “You may qualify” or “You might be eligible” — hedged language, explicitly probabilistic.
- “Estimated benefit” or “Estimated amount” — the word “estimated” is doing real work here; it’s not decorative.
- “Based on the information you provided” — a flag that nothing has been verified, only self-reported.
- “Contact your local office to apply” or “This is not an application” — a direct admission that the process hasn’t started.
- Ranges instead of specific figures — a benefit amount given as a range (“between roughly X and Y”) is a sign the tool is modeling uncertainty rather than calculating a fixed number.
Words that suggest something closer to an actual determination process:
- Language referencing a submitted application, a case number, or an application ID.
- References to specific documents being reviewed or requested.
- A stated timeline for a decision (“you will receive a determination within—check your state’s posted processing time”).
- Any mention of appeal rights or a fair hearing process—these only attach to actual decisions, not estimates.
If a results page gives you a specific dollar figure with no hedging language, no mention of verification, and no next step other than “learn more,” that specificity is often cosmetic. It’s easy to build a calculator that outputs a precise-looking number; it’s much harder to build one that’s actually entitled to that precision given the inputs it collected. Precision in the output doesn’t tell you anything about precision in the underlying calculation—if anything, an oddly exact number from a five-question form should raise your skepticism, not lower it.
It’s also worth noticing who’s running the tool. A screener hosted directly on a state agency’s domain, integrated with their application system, is closer to the determination end of the spectrum—sometimes information you enter there does carry forward into an actual application. A screener hosted by an unaffiliated nonprofit, advocacy group, or commercial site is almost always a pure estimate tool with no connection to any agency’s systems, however well-designed it is. Neither category is inherently more or less accurate, but they carry different weight, and the site’s “About” or “How this works” page will usually tell you which one you’re on if you look for it.
Why Treating an Estimate as Final Leads to Bad Planning Decisions
The practical risk here isn’t abstract. People make real decisions based on screener output, and when the output is treated as more certain than it is, those decisions can go wrong in predictable ways.
Consider a few patterns that show up often:
- Skipping the application entirely. Someone runs a screener, gets a result suggesting they’re likely over the income threshold, and decides not to bother applying—when in reality, a deduction the screener didn’t ask about (child care costs, medical expenses, shelter costs) would have brought them under the line in an actual determination.
- Budgeting around an estimated amount. Someone sees an estimated benefit figure and plans a month’s expenses assuming that number will arrive, only to receive a different amount once household composition and actual documented income are factored in by a caseworker.
- Treating a favorable estimate as a guarantee. A “likely eligible” result gets treated as equivalent to approval, leading someone to make commitments—signing up for a service, assuming coverage will be in place by a certain date—before any actual determination has occurred.
- Assuming one screener’s “no” is the final word. Because different tools model different subsets of the rules, a negative result on one screener doesn’t mean an application would also come back negative. People sometimes stop looking after the first discouraging result.
The common thread in all of these is the same: an estimate is a starting point for deciding whether to invest time in an application, not a substitute for the application’s outcome. It’s useful for the same reason a weather forecast is useful—it helps you decide whether to bring an umbrella—but nobody would rearrange their entire week around a forecast the way they would around a confirmed event on their calendar. Benefit estimates deserve roughly the same posture: informative, worth acting on in the sense of “this is worth applying for,” but not something to build firm financial plans on until an actual agency has looked at actual documents.
If a screener result matters enough to influence a real decision—whether to apply, how to plan a budget, whether to make a commitment that assumes a benefit will come through—the next step is the same regardless of what the tool told you: apply, or contact the relevant agency directly, and treat whatever it produced as a reason to make that call, not as the answer itself.
