ChatGPT Plus on Loper Bright's Superficial Claim of Court's Ability to Determine a Single Best Reading in State of Ambiguity (9/24/26)

In my recent article, John A. Townsend, Loper Bright Flip Flops on Chevron Deference: A Tax Lawyer's Perspective, 79 Tax Law. 323 (2026), on SSRN here, I critiqued the Court’s opinion in Loper Bright Enterprises v. Raimondo, 603 U.S. 369 (2024) (C.J. Roberts), here [Preliminary Print]. One of the points I made was that the Court’s claim that “statutes, no matter how impenetrable, do— in fact, must—have a single, best meaning.” (603 U.S., at 400.) The Court also claimed that courts could and must discern that single best meaning without a default rule to decide the case—such as Chevron and, presumably, some substantive canons that require ambiguity. In Loper Bright’s sweeping claims, no room seemed to be left for statutory ambiguity. However, statutory ambiguity was a common phenomenon before Loper Bright; Loper Bright cannot by fiat eliminate statutory ambiguity.

I decided to use ChatGPT Plus to further explore the phenomenon of statutory ambiguity. That AI tool gave surprisingly cogent answers and step by step reasoning that highlight the lapse in logic of Loper Bright. I wish I had the responses in preparing the article cited above. I offer below the last two prompts and response (after refining earlier prompts and responses). I have left the bold-face and italics as ChatGPT had them; also, ChatGPT sometimes refers to Loper Bright as Loper and does not italicize Loper or Chevron. I have not changed that. Remember these are the final prompts and responses after earlier prompts drilled down to these final two prompt. Those interested in seeing the entire discussion with ChatGPT Plus may do so here.

I am impressed with how ChatGPT responds in an analytically meaningful way to step through  the issues and reach its conclusions. To readers interested in this stuff, ENJOY!

Two points about the Responses: 

1. The first response below refers to my article above without citing the article or otherwise attributing it to me. See par. 4 below. I do not know if ChatGPT Plus recognized that the article that it cited was written by the person with whom it was "chatting." I infer that it did not recognize me as the author of the article.
2. The first response (par. 5) refers to the Flight Options case which it discussed in earlier responses and which I discuss in Interesting Concurring Opinion on Canons (or Maxims) of Statutory Interpretation (Federal Tax Procedure Blog 5/29/26; 8/21/26), here.

JAT Prompt #1: 

In the foregoing discussion, you used 50-50 as the state of interpretive equipoise. Can interpretive equipoise be encountered in a range--say of 45-50 or even 40-60. (These numbers are in percentages.)

ChatGPT Response to Prompt # 1:

Yes—but I would refine the terminology. Interpretive equipoise need not mean a mathematically exact 50–50 tie. It can sensibly describe a zone of interpretive uncertainty within which a court cannot determine, with sufficient confidence, that one interpretation is genuinely “better” than the other. There is, however, no judicially established numerical range such as 45–55 or 40–60.

That distinction becomes important under Loper Bright.

1. Why literal 50–50 is too artificial

Our earlier 50–50 formulation was useful because it made the problem visible:

Agency interpretation: 50%
Opposing interpretation: 50%

Obviously a judge does not actually assign those numbers. And it would be unrealistic to suppose that statutory interpretation permits distinctions as fine as:

A = 50.1%
B = 49.9%

and then requires the court to announce A as the “single, best meaning.”

Interpretation is simply not capable of that degree of measurement. Text, context, structure, linguistic canons, purpose, precedent, and Skidmore considerations do not produce a numerical probability.

Indeed, the Supreme Court's own discussions of ambiguity acknowledge this problem. Justice Kavanaugh has noted that “grievous ambiguity” leaves unresolved the critical question of how much ambiguity counts as enough ambiguity. His Shular concurrence says lenity applies after the traditional tools have been exhausted and the court can do no more than “guess” at statutory meaning.

So I think 50–50 should be understood as the center point of equipoise, not necessarily its entire domain.

2. A 45–55 range seems conceptually quite plausible

Suppose a judge, if forced to quantify matters, said:

Reading A seems 55% plausible;
Reading B seems 45% plausible.

There are two possible ways to understand that statement.

One is:

“I have actually determined that A is more persuasive than B.”

If that is genuinely what the judge means, Loper Bright would seem to say: A is the best reading. Apply A. A modest substantive canon should not displace it merely because the margin is small.

But there is another, more realistic meaning:

“My best rough impression is 55–45, but the interpretive materials are not sufficiently discriminating for me confidently to say that A is legally superior to B.”

That is quite different. In that situation, the 55–45 estimate is really an expression of uncertainty, not a finding that A has been established as the best reading.

I would regard that as interpretive equipoise.

3. The same could be true even at 60–40

A 60–40 split presents a harder case, but I do not think it can categorically be excluded.

Again, everything depends upon what “60–40” means.

If the judge says:

“After considering everything, I am reasonably confident A is superior to B; if I had to put numbers on it, perhaps 60–40,”

then there is no equipoise. The court has identified its best reading.

But suppose the judge says:

“There are substantial considerations favoring each interpretation. I have a weak inclination toward A, perhaps 60–40, but I cannot say with sufficient confidence that the interpretive tools establish A rather than B.”

Then one might reasonably characterize the case as within an interpretive-equipoise zone.

The question therefore is not really:

Is the number 55 rather than 50?

It is:

Has the interpretive process produced a sufficiently reliable basis for distinguishing one interpretation as the best?

That is a much better test.

4. There is now a published 40–60 formulation of exactly this idea

Interestingly, a recent discussion in The Tax Lawyer expressly develops the probabilistic model you describe. It proposes an illustrative 40%–60% range as an area of interpretive equipoise in which the court cannot determine with sufficient confidence which interpretation is “best” or “not best.” It also makes the point that meaningful judicial discrimination between, for example, 51–49 interpretations is unrealistic and suggests that 45–55 or 40–60 might better represent the practical zone of uncertainty.

That is an analytical model, not a doctrinal numerical rule. But I think it captures something important that the simple 50–50 example misses.

5. Flight Options supports a range concept, although it does not quantify it

Judge Murphy's concurrence in Flight Options describes a modest canon as operating between “equally plausible interpretations” and says the pro-taxpayer canon has this limited tie-breaking function. He also analogizes it to lenity, which may operate when “reasonable doubts” concerning the best reading persist after the traditional tools have been exhausted.

I would not read “equally plausible” as requiring a judge to conclude:

A = precisely 50.000%
B = precisely 50.000%.

That would make the canon almost nonexistent.

“Equally plausible” is better understood qualitatively:

Neither interpretation has emerged as sufficiently superior to the other to permit the court confidently to identify it as the best reading.

That describes a range rather than a mathematical point.

6. This is analogous—but not identical—to burdens of persuasion

There is a useful analogy to factfinding.

We commonly say that a civil plaintiff must prove something by a “preponderance”—more likely than not. People sometimes translate that into 51%.

But real factfinders do not calculate:

50.00001% = proven;
49.99999% = not proven.

The verbal standard identifies a degree of confidence, not an actual mathematical computation.

Interpretive confidence is even less susceptible to quantification because there is no empirical frequency from which probabilities can readily be derived.

So percentages are useful for modeling the problem, but should not be mistaken for actual judicial measurements.

7. I would distinguish “strict equipoise” from an “equipoise zone”

This may give you the cleanest terminology.

Strict equipoise:

A and B appear exactly equally persuasive—the conceptual 50–50 case.

Interpretive-equipoise zone:

A range surrounding strict equipoise in which the differences between A and B are too uncertain or insubstantial for the court responsibly to identify either as the best interpretation.

Thus one might illustrate:

Illustrative assessment	Characterization
50–50	Paradigm equipoise
52–48	Almost certainly within practical equipoise
55–45	Quite plausibly within an equipoise zone
60–40	Could be, depending on confidence in the distinction
70–30	Ordinarily indicates a best reading
90–10	Clearly not equipoise

I emphasize that those numbers illustrate a concept; they are not legal thresholds.

8. This exposes a difficulty in Loper's “best reading” formulation

Taken literally, Loper Bright might seem to say:

If A is even marginally more persuasive than B, A is the “single, best meaning.”

But that formulation assumes a degree of interpretive precision that courts cannot realistically achieve.

Suppose:

A appears 51% persuasive.
B appears 49% persuasive.

Calling A the legally ascertainable “best meaning” implies that the judicial process can reliably distinguish those two levels of persuasiveness.

It probably cannot.

This is precisely the longstanding “ambiguity about ambiguity” problem. Farnsworth, Guzior, and Malani's empirical work showed that judgments about whether statutory language is ambiguous themselves vary considerably and can be affected by how the ambiguity question is framed. And before Loper, Justice Gorsuch criticized Chevron partly because no coherent answer had emerged to the question of how much ambiguity is enough to trigger deference.

Loper eliminates Chevron but does not eliminate that epistemological problem.

9. I therefore would formulate the test without percentages

For legal analysis, I think this formulation is stronger than saying “40–60”:

Interpretive equipoise exists when, after exhausting the relevant tools of statutory interpretation, the court lacks a sufficiently reliable basis for concluding that one competing reasonable interpretation is superior to the other. Equipoise therefore describes a zone of irreducible interpretive uncertainty rather than necessarily an exact balance between two interpretations.

Then you could add in a footnote:

The concept may be illustrated probabilistically as a range centered on 50–50—perhaps 45–55 or 40–60—but the percentages are heuristic rather than juridical thresholds.

That avoids an obvious criticism: if you define “best” as anything above 50%, critics can simply respond that a 60% interpretation is necessarily “better” than a 40% interpretation.

Your stronger response is that the court cannot actually know that its nominal 60–40 impression is sufficiently reliable to constitute a legal determination of superiority.

10. That matters directly to the tax-canon question

It also improves our preceding discussion of the pro-taxpayer canon.

The canon need not wait for the practically impossible finding:

“These interpretations are exactly 50–50.”

Instead, it can operate when:

ordinary interpretation has narrowed matters to two reasonable readings, but has not supplied a sufficiently reliable basis for identifying either as the superior reading.

At that point, a modest canon supplies the constructional rule:

unresolved doubt → taxpayer-favoring construction.

This seems to me the most defensible understanding of Judge Murphy's statement in Flight Options that the canon operates between “equally plausible” readings.

So yes, I think your 40–60 conception is analytically defensible, provided “40–60” denotes a zone in which the court cannot reliably discriminate between the competing interpretations, rather than a literal finding that one interpretation has a demonstrably 60% probability of being correct. That qualification is crucial.

JAT Prompt #2: 

If there is a zone in which the court cannot reliably discriminate between the competing interpretations, how does the court make the decision as to which interpretation to apply (assume that there are no canon that act as tie-breaker)?

ChatGPT Response #2:

Under Loper Bright as written, the court is supposed to choose one interpretation as the “best” reading even if its confidence is low. The majority does not recognize a residual category in which the court may say, “the interpretations are indistinguishable, so I need some noninterpretive default.” It insists that the statute has a “single, best meaning” and that the court must use every tool available to identify it.

But if we accept your premise—all interpretive tools have been exhausted, no tie-breaking canon applies, no delegation exists, and the court genuinely cannot reliably discriminate between A and B—then I think Loper contains a real logical gap. It tells the court that there is a best answer, but not how to choose it when the court cannot know which answer is best.

1. Loper's formal answer: keep interpreting until one reading is “best”

The majority is unusually categorical. It says that statutes “no matter how impenetrable” must have a “single, best meaning,” and that courts should “use every tool at their disposal” to resolve ambiguity. It further says that in an agency case the governing interpretation is simply “the reading the court would have reached” had no agency been involved.

That implies something important for our earlier discussion of an “equipoise zone.”

Strictly speaking, the Loper majority does not accept such a zone as a permanent endpoint of interpretation. The court may have difficulty, and the competing readings may be close, but ultimately the judge is supposed to say:

A is better than B.

Even if the judge's confidence is modest.

Thus, if the judge thinks A is actually somewhat more persuasive—our hypothetical 55–45 or 60–40—and believes that distinction is reliable, Loper says apply A.

2. But your hypothetical is different

You are stipulating something stronger:

The court has exhausted everything legitimately bearing on meaning and concludes that it cannot reliably distinguish A from B.

So this is not:

A is slightly better than B.

It is:

I cannot responsibly say whether A or B is better.

That creates an epistemological problem that the statement “there must be a best meaning” does not solve.

There may objectively be a correct historical fact about what happened in a room 100 years ago. But if the evidence is perfectly balanced, telling the factfinder that “one thing really did happen” does not tell the factfinder which thing to find.

The same is true here:

Ontological claim: The statute has one correct meaning.

does not entail:

Epistemological claim: This court can determine which meaning is correct.

Loper largely moves from the first proposition to the second without explaining that gap.

3. Justice Kagan's dissent identifies essentially this problem

Justice Kagan rejects the majority's premise. She argues that sometimes a statute genuinely has multiple reasonable readings and “Congress has not chosen among them.” Her question is then: who decides which reading governs? Chevron answered: within its domain, the agency. The majority instead says the court can always find the best interpretation.

The disagreement is therefore deeper than agency deference.

The majority's position is essentially:

Interpretation always produces an answer.

The dissent's position is:

Sometimes interpretation runs out before an answer has been produced.

Your hypothetical assumes that the dissent is descriptively correct on that point, even though Chevron is no longer the rule.

And then the question becomes: what replaces Chevron's decision rule?

Loper does not say.

4. A court cannot simply refuse to decide

Of course, the judge cannot write:

“A and B are indistinguishable. Judgment withheld indefinitely.”

A concrete lawsuit requires a disposition.

But there is an important distinction between:

deciding what the statute means, and

deciding which party receives judgment when statutory meaning cannot be ascertained.

The latter requires a rule of decision.

For example, if lenity applied:

unresolved ambiguity → defendant wins.

If the Gould/Merriam tax canon applied:

unresolved ambiguity → taxpayer wins.

If Congress had delegated the choice:

unresolved statutory specification → agency may choose within delegated limits.

But your hypothetical removes all those possibilities.

Now there is no identified legal rule connecting:

A = B

to:

choose A rather than B.

That is the missing step.

5. Could the court just choose whichever reading seems microscopically better?

There are really two possibilities.

Suppose the judge says:

“I cannot be highly confident, but on balance A seems marginally preferable.”

Then Loper gives an answer:

Choose A.

The relevant standard is not “beyond reasonable doubt,” “clear and convincing evidence,” or even formally “preponderance.” Loper simply asks for the best reading. So the doctrine seemingly permits a very low-confidence determination that one interpretation is better.

But that is not your hypothetical.

You posit that the judge says:

“Any apparent preference for A over B is within the noise of the interpretive process. I have no reliable basis for saying A is better.”

At that point, telling the judge to select “the best” interpretation is circular:

Choose the better interpretation.
Which is better?
I cannot determine that.
Choose the better one anyway.

That is not a decision rule.

6. This is why the “equipoise zone” is analytically important

I would now formulate the problem this way.

There are three rather than two categories:

Category 1 — Determinate superiority

A is demonstrably better than B.

Apply A under Loper.

Category 2 — Close but discriminable

A is only somewhat better than B, but the court believes the difference is real and reliable.

Still apply A under Loper. The Court does not impose a minimum margin of superiority.

Category 3 — Nondiscriminable

The court cannot reliably determine whether A or B is superior.

That is true interpretive equipoise.

And Loper provides no express rule for Category 3, because its theoretical premise is that Category 3 need never be the endpoint of a properly conducted interpretation.

That is why CRS has observed that the majority's “single best meaning” conception sits in tension with the continuing reality of statutory ambiguity.

7. Could ordinary litigation burdens supply the missing rule?

Potentially they could decide the judgment, but—as we discussed earlier—that does not necessarily determine the meaning of the statute.

Suppose a refund plaintiff needs Interpretation T to recover.

If T and G remain nondiscriminable, one might say:

Taxpayer has not established entitlement to recovery → Government gets judgment.

But notice what has happened.

The court has not interpreted the statute as meaning G.

It has instead said:

I cannot determine whether T or G is correct, but a procedural or substantive rule allocates the consequences of uncertainty against this litigant.

That is a rule of decision, not an interpretation.

And, as our circuit survey indicated, the taxpayer's ordinary evidentiary burden is not clearly established as such a default for pure questions of statutory law.

The same distinction would apply in non-tax litigation.

8. Flight Options actually highlights the problem

Judge Murphy's concurrence in Flight Options is useful because it says that “modest” substantive canons operate as tie-breakers between “equally plausible interpretations,” and specifically places the pro-taxpayer canon in that category.

That makes perfect conceptual sense:

ordinary interpretation → equally plausible readings → modest canon → answer.

But it also allows us to see what happens when your hypothetical removes the last step:

ordinary interpretation → equally plausible readings → no applicable canon → ?

Neither Flight Options nor Loper supplies a general “?” rule.

9. There are therefore two different descriptions of what a court would actually do

The doctrinal description under Loper would probably be:

The court continues exercising independent judgment and eventually declares one interpretation the best reading.

Courts are unlikely to announce that they have reached irresolvable equipoise, because Loper tells them that statutory ambiguity is ultimately resolvable through judicial interpretation. One appellate commentary accurately summarizes Loper as giving courts no Chevron-style “off-ramp or tie breaker”; the court simply must determine the best reading.

The jurisprudential description, however, may be different:

If the interpretive evidence really is nondiscriminating, the judge's final selection necessarily rests on something other than superior evidence of statutory meaning—perhaps intuition, background assumptions, an unstated substantive canon, a presumption, institutional considerations, or simply a judgment call.

Calling the resulting interpretation “best” does not itself identify the basis on which the tie was broken.

10. That produces a somewhat paradoxical result

Assume:

A = 50
B = 50

and absolutely everything else has been stipulated away.

The court must choose.

Suppose it chooses A.

There are only three logical possibilities:

1. The court discovered some additional reason favoring A.
   Then there was not really equipoise.
2. The court applied some legal default favoring A.
   Then there was a tie-breaker, contrary to our assumption.
3. The court simply selected A without a reason capable of distinguishing A from B.
   Then the selection is not really an act of interpretation based on superior statutory meaning.

There isn't a fourth possibility.

That is why I think your hypothetical exposes an important weakness in the strongest reading of Loper's “single best meaning” proposition.

My answer in its shortest form

If true interpretive equipoise persists and there is no applicable canon, delegation, presumption, or other rule of decision, then:

Loper Bright does not provide a principled tie-breaking rule. It instead assumes that the court can and must resolve the ambiguity by identifying a best reading. If the competing interpretations genuinely cannot be reliably distinguished, that assumption fails, and some extra-interpretive rule of decision is logically necessary even though Loper does not identify one.

I would therefore be careful about saying that the court should simply pick whichever interpretation is “51% better.” That solves the problem only if the court actually believes that the 51–49 difference is reliably meaningful. If the court's conclusion is instead that the difference cannot reliably be discerned, declaring one interpretation “best” merely labels the result; it does not explain how the result was reached.

And that, I think, may be the central theoretical vulnerability in Loper Bright when one takes its “single, best meaning” language literally.

JAT Postscript: If you can accept that Loper Bright has not taken away the potential for statutory ambiguity, then I urge readers to consider how a court decides the case in the state of ambiguity /aka equipoise (assuming no other rule or canon of construction supplies a default rule like lenity). In an agency interpretation case (assume IRS), should the taxpayer win or should the IRS win? The court has to decide a winner or loser on the interpretive issue. Further, is the default in favor of the IRS under (i) APA § 706(2)(A)'s requirement that the court determine the agency interpretation is "not in accordance with law" where the court cannot so determine or (ii) does the ordinary burden of persuasion supply the default rule because the taxpayer has not persuaded that he should prevail. And, after a lower court (say district court) decides the case in equipoise and the appellate court is in ambiguity, does it review the lower court's "interpretation" de novo where it applies its own hunches/guesses in ambiguity and reverse or affirm the lower court accordingly?

My conclusion: this is just one instance where Loper Bright is a mess.