What is plausible distortion?
Plausible distortion is what happens when an AI output looks correct but has changed the meaning of the source material decisions depend on. Each place the meaning moved is a shift. The output reads fluently, passes casual review, and reaches the decision-maker looking accurate, while the signal has been softened, compressed, omitted, or redirected somewhere between the source and the output.
The output that looked right
Hallucination is one failure class: outputs that contain claims with no basis in any source. Plausible distortion is another, and the two ask different questions. One asks whether a claim has source support. The other asks whether the source's meaning survived.
Plausible distortion is difficult to notice because the output can remain closely connected to the source. The problem is not absence of support; it is that important meaning moved while the output remained credible. The output stays close to the source. It uses the right vocabulary, covers the right topics, and reads like a faithful summary. But somewhere in that gap between source and output, meaning changed. A condition was dropped. A risk was softened. A qualifier disappeared. A conclusion was drawn that the source never actually reached.
The output was not wrong enough to trigger review. That is exactly when the damage happens.
AI sits between the source and the person reasoning from it. What survived the output is not always what the source actually contained. Plausible distortion is the gap between the two.
Signal means the decision-relevant meaning in the source: risk, urgency, uncertainty, conditions, weighting, and context. Plausible distortion is what happens when signal changes or disappears between the source and the output.
A concrete example
Here is what plausible distortion looks like in practice. Source material from a vendor contract, and the AI output produced from it:
The vendor is obligated to deliver a remediation plan within 30 days of any material breach, subject to written approval from both parties before implementation.
The vendor is expected to deliver a remediation plan within 30 days of any material breach.
Changed. "Obligated" softened to "expected." The written approval requirement was dropped entirely. The output reads correctly. The meaning changed.
Nothing in this output is invented. The vendor, the 30-day window, the remediation plan are all present and all accurate. But "obligated" and "expected" do not mean the same thing in a contract. And the approval requirement was decision-relevant. It disappeared with no signal that it was ever there.
The forms it takes
Plausible distortion is not a single failure mode. Plumb traces movement across the dimensions that carry meaning: conditions, exceptions, certainty, scope, obligations, permissions, timing, numbers, entities, relationships, causality, and emphasis. Some of the forms that movement takes:
The signal got quieter
Source material is present in the output but its meaning has shifted. A risk is softened. An obligation becomes a suggestion. Uncertainty is resolved into a cleaner claim than the source justified.
The source had more
Source units with material signal are absent from the output. Nothing in the output signals the gap. The output looks complete. It is not.
The inference outran the source
The output presents a conclusion or recommendation as if the source directly supported it. Synthesis is not the problem. Unmarked synthesis is.
No source trace at all
Claims that cannot be traced to the source. This is the hallucination case, and the one grounding and hallucination evaluation most directly address.
Grounding, faithfulness and hallucination-evaluation methods can detect several kinds of source and output failure, including unsupported claims and some semantic inconsistency. They do not by themselves guarantee that every material condition, exception, qualifier or source relationship survived the transformation. Why a supported claim can still get the source wrong works through that distinction in detail.
Why it is difficult to catch
An unsupported claim gives you something to challenge: you can ask where it came from, and the answer is nowhere.
Plausible distortion is different because the claim does have a source. The source just said something slightly different. "Obligated" became "expected." A 30-day hard deadline became a general expectation. A conditional approval requirement disappeared. Each change is subtle enough to pass a reader who is not comparing the output directly against the source, which is most readers, most of the time.
That is the problem. Not that the output was wrong. That it was wrong in a way that looked right.
Why better prompts do not solve it
A common response to AI output quality problems is to improve the prompt. Use clearer instructions. Add constraints. Tell the model to preserve exact language. These steps reduce error frequency but do not eliminate plausible distortion.
Better prompting can reduce these failures, but a strongly prompted output can still omit a condition, change certainty or lose a relevant qualification. The model is not breaking the instruction. It is producing a fluent account of the source rather than a transcription of it.
Checking the output directly against the bounded source makes those changes inspectable.
How Plumb Analysis detects it
Plumb takes two things: the material the AI worked from, and the AI output exactly as produced. It breaks both into comparable pieces of meaning, then reads the output against the source across the dimensions that carry meaning: conditions, exceptions, certainty, scope, obligations, permissions, timing, numbers, entities, relationships, causality, and emphasis.
For every claim, it shows what changed, where it changed, and the source language behind the finding:
The result can be attached to the output before it reaches the decision-maker, so the person acting on it sees what it was actually entitled to claim.
Is plausible distortion the same as hallucination?
No. Hallucination is when an AI output contains claims that have no basis in any source. Plausible distortion is when the output stays close to the source but changes its meaning: softening a condition, dropping a qualifier, omitting a risk, or redirecting a conclusion. The two ask different questions: one is about claim support, the other about whether the source's meaning survived.
Where does plausible distortion show up?
Anywhere an AI output is produced from a known set of material and then acted on. Contract review. Investment memos. Consulting deliverables. Project briefs. Regulatory summaries. Risk assessments. The common thread is that someone reads the output and makes a decision without going back to the source to check what it actually said.
Can you catch it manually?
Sometimes. If you are comparing a short output against a short source, a careful reader can catch changed meaning. But manual review does not scale, does not apply consistently, and does not catch omissions, because you cannot notice what is missing from an output unless you have already read the source. Most of the time, people read the output, not the source. That is the point of an AI output.
Is this a problem with specific AI models?
Not to one model. Language models can transform source meaning while summarizing, synthesizing and generating from source material. Better prompting, retrieval and model capability can reduce these failures, but they do not guarantee preservation of every condition, qualifier, exception or relationship. Different models have different error profiles.
Is plausible distortion the same as AI drift?
They describe the same movement from different angles. AI drift is the general case of meaning moving between a source and an output. Plausible distortion is the case where that movement leaves the output looking correct.
What is Plumb Analysis?
Plumb Analysis reads AI-generated output against the bounded information behind it and surfaces where meaning changed, disappeared or was introduced. It does not generate, summarize or score. Plumb Boundary is the companion product, for what AI does rather than what it produces. Plumb is built by Astrelle Intelligence Inc., based in Toronto.
See what your AI outputs actually said
The demo takes 15 minutes. We run Plumb against a real output, yours or one of ours, and show you exactly what it finds.
Talk to Us →The products
Plumb reads what AI produced against the information behind it, and what AI did against what actually became true.
Plumb Analysis for what AI produces · Plumb Boundary for what AI does