Press

AI Slop Check

A short form that builds a document-review prompt you can paste into any assistant. It scores how much of a document is specific, evidenced, and worth a colleague's time, and how much of it could have been written for any company.

Why this scoring exists

It exists to protect the time, attention, and judgment of the people you work with. A document is not ready to circulate just because it has been written. Before you ask anyone to spend time on a document, you owe them a document worth their time. The score is a way to check that before the cost lands on someone else.

Respect for colleagues

Every unnecessary paragraph, vague claim, repeated explanation, or generic section is a cost the reader pays in time, confusion, follow-up questions, meetings, and rework. The writer should pay the cost of fixing the document before handing that cost to everyone else.

Ownership

The person who creates or sponsors a document owns its quality. AI can help draft, critique, shorten, and test it, but AI does not take over ownership. A high score means the document is not ready to become someone else's problem, and the writer keeps working until it is specific, supported, and useful.

Understanding shows in writing

Writing is evidence of understanding. Someone who understands a problem can say what happened, what is known, what is not known, why it matters, what supports the conclusion, and what should happen next. Generic language and vague recommendations often mean the writer has not understood the subject deeply enough yet, so the score measures demonstrated understanding, not polish.

The quality of organizational thinking

Important decisions should not rest on anecdotes and differently remembered conversations. A strong document is a durable record of what the organization believes to be true, one that colleagues can challenge, correct, assign, verify, and learn from later. The score checks whether the record has enough substance to support that.

The standard

The goal is not shorter documents. The goal is documents that earn the reader's time. A long document full of necessary evidence and reasoning can be excellent. A short one full of vague statements can still be slop. The rule is minimum words necessary to preserve maximum useful information.

Treat the score as a gate before a document circulates. High score, keep working. Low score, the document has earned its readers. And the question underneath every review is one sentence: could large parts of this be dropped into another company's document unchanged?

Run the check

Answer what you can and leave the rest blank. The prompt marks blank fields as not provided instead of guessing. Paste the document itself into the last field, or paste it into your assistant after the prompt.

How to read the score

0-1: Exceptional. Almost every sentence carries necessary company-specific information, evidence, reasoning, or action.
2-3: Strong. Minor unnecessary material, but specific and useful throughout.
4-5: Acceptable but needs editing. Real content is present, buried under generic language, repetition, unsupported claims, or background the reader already knows.
6-7: High slop. The useful information is inside substantial generic or repetitive content.
8-9: Very high slop. It sounds professional and says little that is specific enough to act on.
10: Essentially content-free. It could have been written for almost any company with only the names changed.

How the score is weighted

Company, customer, and market specificity, 30%
Evidence and factual grounding, 20%
Actionability and operational usefulness, 20%
Repetition and unnecessary content, 15%
Precision of language, 10%
AI-shaped formatting and rhetorical filler, 5%

Nothing entered here is stored or sent. Review the prompt before sharing confidential, personal, patient, or privileged information.

Document typeWhat kind of document is it? The reviewer checks every section against this purpose.The company and what it actually doesName it and say what it sells, builds, or operates. Two or three plain sentences beat a mission statement.The business unit or team this document comes fromWho wrote it and who they write for inside the company.The market it operates inIndustry, competitors, and the regulatory, reimbursement, or payment realities that change how the document should be read.The customer and their workflowWhich customer type, what they are trying to do, and what a bad outcome looks like for them.Who reads this document and what they must decideThe actual audience, and the decision, approval, or action the document is asking of them.Evidence already in the documentDates, incidents, tickets, logs, metrics, messages, or source documents the writer cites. List what is there, not what should be.The document itselfPaste the full text here. You can also leave this blank and paste the document into your assistant after the prompt instead.

The full review: the score, the sources of slop, the cuts, and the header block.

Review the prompt

AI SLOP CHECK

You are reviewing a business or operational document for quality, specificity, evidence, and usefulness. Your job is not to judge whether the writing sounds polished. Your job is to determine how much of the document is useful, specific, evidence-based, operationally relevant content, and how much is generic, repetitive, vague, inflated, or unsupported material that could have been written for almost any company.

Produce an AI Slop Score from 0 to 10. 0 means exceptionally specific, concise, evidence-based, and grounded in this company, its customers, its market, and the issue at hand. 10 means the document could have been written for almost any company with only the names changed.

The central test: could substantial portions of this document be copied into a document for another company without changing the substance? If yes, the score goes up.

The goal is minimum words necessary to preserve maximum useful information.

CONTEXT
The company and what it actually does: Not provided
The business unit or team this document comes from: Not provided
The market it operates in: Not provided
The customer and their workflow: Not provided
Who reads this document and what they must decide: Not provided
Evidence already in the document: Not provided
Document type: Not provided

TASK
Review the document and produce the full scored output.

Evaluate the document on these criteria.

1. Company specificity. Look for concrete references to the actual company: products, services, systems, workflows, teams, roles, organizational structure, technologies, processes, constraints, actual incidents, actual decisions, actual operating practices. Ask of each paragraph: what here could only have been written after understanding this company? If the answer is nothing, treat it as potentially high-slop.

2. Customer specificity. The document should reflect actual customer realities, not vague references to customers, users, or stakeholders. Look for specific customer types, workflows, environments, expectations, impact, incidents, support issues, implementation constraints, and integration requirements. A sentence like "this negatively impacted the customer experience" is slop unless the impact is explained.

3. Market specificity. The document should reflect the realities of the company's actual market: industry, regulatory environment, reimbursement or payment models, competitors, market structure, buying behavior, operating constraints, standards, technologies, and industry-specific risks. Do not reward jargon. Reward information that changes the reasoning because this company operates in this market.

4. Evidence versus assertion. Identify consequential claims presented as fact without sufficient evidence. Evidence means dates, incidents, tickets, emails, messages, logs, metrics, customer communications, system behavior, timelines, examples, and source documents. Unsupported conclusions raise the score.

5. Actionability. A reader should know what happens next: owner, action, expected result, deadline or trigger, evidence required, decision required, acceptance criteria, escalation path, follow-up, verification. "Improve communication," "monitor the situation," "ensure better testing," and "implement appropriate controls" are intentions, not instructions. Penalize them.

6. Repetition. Penalize the same idea restated in different words: summaries that merely repeat prior sections, conclusions that repeat recommendations, and redundant bullet lists. Do not penalize repetition that serves a clear operational purpose.

7. Unnecessary background. Penalize material the intended reader already knows that does not contribute to a decision, investigation, action, or understanding. Generic explanations of common business, engineering, management, or industry concepts should usually be cut.

8. Vague language. Flag words like robust, comprehensive, seamless, effective, efficient, appropriate, strategic, proactive, optimized, best practices, industry standard, significant, various, numerous, stakeholders, and leverage. They are not automatically bad; they are slop when they substitute for concrete information.

9. AI-shaped writing. Do not penalize a document because AI helped write it. Penalize the patterns: excessive headings and bullet lists, artificial symmetry, repetitive summaries, generic executive summaries, unnecessary definitions, filler introductions, conclusions that restate the document, ceremonial language, and polished wording that carries little information. The test is usefulness, not whether AI was involved.

10. Information density. Ask: if 25% of the words were removed, would anything important disappear? Then ask the same for 50%. Estimate how much could be removed while preserving every meaningful fact, piece of evidence, reasoning step, decision, instruction, responsibility, and acceptance criterion. The larger the removable share, the higher the score.

11. Operational truth. The document must separate verified facts, reported observations, interpretations, hypotheses, conclusions, and unresolved questions. Penalize documents that turn assumptions, opinions, or anecdotes into facts.

12. Purpose alignment. Infer the document's purpose and judge whether every major section serves it. Content can be valid and still be slop if it does not belong in this particular document.

Scoring guidance:
0-1 Exceptional. Almost every sentence contains necessary company-specific information, evidence, reasoning, or action.
2-3 Strong. Minor unnecessary material, but highly specific and useful.
4-5 Acceptable but needs editing. Meaningful content is present, but generic language, repetition, unsupported assertions, or unnecessary explanation should be reduced.
6-7 High slop. Useful information is buried inside substantial generic or repetitive content.
8-9 Very high slop. The document sounds professional but contains little information specific enough to guide the company.
10 Essentially content-free. It could have been written for almost any company with only names changed.

Weight the score approximately: company, customer, and market specificity 30%; evidence and factual grounding 20%; actionability and operational usefulness 20%; repetition and unnecessary content 15%; precision of language 10%; AI-shaped formatting and rhetorical filler 5%.

A polished document written for "a company," "a SaaS company," "an engineering organization," or any broad category should not score well. The standard is company-specific usefulness. Do not reward length. Do not reward brevity. Do not reward polish. Reward specificity, evidence, reasoning, and operational usefulness.

Required output:

AI SLOP SCORE: X.X / 10

VERDICT: one sentence stating whether the document is ready to circulate.

COMPANY SPECIFICITY SCORE: X / 10. 10 means deeply grounded in the actual company, customers, market, systems, and circumstances. 0 means essentially interchangeable with another company.

ESTIMATED REMOVABLE CONTENT: X%. How much could be removed without losing meaningful facts, evidence, reasoning, decisions, or instructions.

TOP SOURCES OF SLOP: no more than five. For each, name the offending section, explain why it is slop, and recommend delete, shorten, make specific, or support with evidence.

GENERIC CONTENT TEST: identify passages that could appear unchanged in another company's document. Be aggressive.

MISSING COMPANY GROUNDING: the information needed to make the document specific to the company, the customer, the incident or issue, the market, the systems, the workflow, and the people or roles involved. Do not invent missing facts.

UNSUPPORTED ASSERTIONS: consequential claims the document does not support with sufficient evidence.

QUESTIONS THE DOCUMENT SHOULD ANSWER: only questions that would materially improve the document. No generic questions.

SURGICAL CUTS: sentences, paragraphs, headings, or sections that should simply be removed.

HIGH-VALUE CONTENT: the strongest portions, containing useful facts, evidence, reasoning, decisions, or instructions.

FINAL EDITING INSTRUCTION: one concise instruction to the writer for the next revision.

Then produce this compact block for the top of the document:

AI SLOP SCORE: X.X / 10
COMPANY SPECIFICITY: X / 10
ESTIMATED REMOVABLE CONTENT: X%
AI QUALITY CHECK: COMPLETED

DOCUMENT UNDER REVIEW
Not provided here. I will paste the document after this prompt. Wait for it before scoring.

You can leave any field blank. The prompt will mark it as not provided.

If a document you care about keeps scoring high and you cannot say why, send us the score and the kind of document it was.

Write to the Press Intellectual Frontiers on LinkedIn