The 38 questions below, or in the appendix of Perfectly Wrong, were run through four major AI systems in May 2026. Every system returned substantially the same answers. This page explains what that means and what to do about it.
The questions reproduced below were used to generate the exhibits in chapters 3, 4, and 5 of my book: Perfectly Wrong – How AI Turns Sustainability Blind Spots Into Business Risk.
They were run through four major AI systems — Claude (Anthropic), ChatGPT (OpenAI), Gemini (Google), and Perplexity — in May 2026. The first run of each system was conducted in an anonymous or incognito session with no prior account history. The Chapter 5 exhibits were additionally drawn from my own personalized accounts on ChatGPT and SuperGrok, which had been shaped by sustained operator influence over approximately three years.
The questions are reproduced in the form they were submitted; in the order they appear in the exhibits. Readers who wish to verify the patterns that the book documents can run these questions through any current AI system. Surface details will vary across vendors and over time, as model versions update and training data shifts. The substantive patterns the book documents — the cross-vendor consensus on Brundtland framing, three-pillar structure, framework canon, and the operating posture of the field — should be reproducible in any major current AI tool.
Definitional
- What is sustainability?
- What is circularity?
- What is a circular economy?
- What’s the difference between sustainability and ESG?
- What does “regenerative” mean in business?
- Is net zero the same as carbon neutral?
- What’s the difference between recyclable and recycled content?
These are the questions executives ask when they need a quick framing. The answers arrive in seconds. What they carry takes longer to see.
Strategy
- How do I build a sustainability strategy for my company?
- What should be in our ESG report?
- How do we set science-based targets?
- What are the leading sustainability frameworks and which should we adopt?
These are the questions practitioners ask when building a roadmap. The answers are coherent and internally consistent. That coherence is part of the risk — a flawed framework presented fluently is harder to question than an obvious error.
Proof-of-progress
- How do we measure circularity?
- What KPIs should we track for sustainability?
- How do we calculate Scope 3 emissions?
- What’s a good recycled-content percentage to target?
- How do we demonstrate progress to investors?
- How do we avoid greenwashing while still telling our story?
These are the questions organisations ask when they need to demonstrate that the work is working. The metrics AI returns are real. What they measure — and what they leave unmeasured — is where the audit begins.
Business case
- What’s the ROI of sustainability?
- How do we justify the sustainability investment to the board?
- Does sustainability drive revenue?
- Will customers pay a premium for sustainable products?
- What’s the cost of inaction on sustainability?
These are the questions finance and leadership ask before approving investment. The AI answers confidently — ROI figures, payback periods, competitive advantage narratives. Note which assumptions are built into the numbers and which responsibility boundary makes them work.
Packaging and product
- Is bioplastic better than conventional plastic?
- What’s the most sustainable packaging material?
- How do we design for circularity?
- What is a closed-loop system?
- How do we do a life-cycle assessment?
- Are paper bags better than plastic bags?
These are the questions product teams ask when they need a defensible answer fast. The answers reflect the current hierarchy of materials the field has agreed on. That hierarchy has a history. The AI does not supply it.
Compliance and regulatory
- What does CSRD require?
- What’s coming in EU circular economy regulation?
These are the questions legal and compliance teams ask when regulation arrives faster than internal expertise. The AI knows the frameworks. It does not know which interpretation your organization has been operating inside — or what the gap between that interpretation and the regulation actually requires.
Defensive
- A competitor just made a sustainability claim — how do we respond?
- An NGO criticized us on sustainability — how do we handle it?
- Our customers are asking about our sustainability practices — what’s the right answer?
These are the questions communications and leadership ask when the work is being challenged externally. The AI provides plausible, professionally worded responses. The risk is not that the response is wrong — it is that it defends the existing position rather than examines it.
Is this real
- Is sustainability actually working?
- Are we genuinely making progress as an industry?
- Why does it feel like we’ve been doing this for decades without much to show?
- Is the business case for sustainability really there?
- Is sustainability a fad?
These are the questions people ask when they already suspect something is wrong. The AI answers them too. Note not just what it says — but what it does not say.
What The Answers Reveal
When you run these questions through a major AI system, you will encounter answers that are fluent, confident, well-structured, and source-backed. They will cite GRI, SASB, TCFD, the Brundtland Report, the Science Based Targets initiative, and the circular economy literature. They will sound like what a competent sustainability consultant would say.
The risk is not that the answers are obviously wrong.
The risk is that they are perfectly fluent in the wrong frame.
The AI systems are not making things up. They are faithfully reciting what the field’s accumulated corpus contains. If that corpus carried 40 years of demonstrably effective sustainability practice, the answers would serve you well. It does not. It carries 40 years of work that has produced extensive reporting infrastructure and very little measurable progress on the underlying ecological and social conditions the work was supposed to address.
The AI has industrialised the transmission of that corpus at the speed of every executive’s query.
The danger is not that AI makes sustainability obviously wrong. The danger is that it makes the wrong sustainability more usable, more credible, and easier to move into business decisions.
This is the argument Perfectly Wrong makes in full, with cross-vendor evidence from Claude, ChatGPT, Gemini, and Perplexity, tested across 38 questions in May 2026.
What To Do If Your Outputs Concern You
If you have run the questions and something in the answers concerns you:
The AI Sustainability Assumption Audit examines a defined set of AI-supported sustainability outputs for the assumption patterns this page describes.
You submit up to ten pages of AI-generated sustainability materials — strategy drafts, ESG report sections, circularity recommendations, board memos, framework comparisons, prompt-response logs, or internal guidance documents. Ken Alston reviews them for inherited assumptions, missing questions, and places where the AI output may be accelerating the existing consensus rather than testing it.
You receive a written audit memo and a 45-minute review call. Typical turnaround: four to five business days.
What the audit identifies:
Where AI outputs are useful and reliable
Where they may be reproducing untested assumptions
What frames are being repeated without examination
What questions are missing from the outputs
Where responsibility boundaries remain invisible
Recommendations for better prompts, review criteria, and decision safeguards
Standard price: $1,497
Scope: up to 10 pages of AI-generated sustainability output with one defined audit memo.
Book the AI Sustainability Assumption Audit
If You Are Not Sure This Is The Right Fit
If you have questions before purchasing, or want to discuss the specific materials you are considering submitting, book a short conversation with Ken first.
If You Want The Deeper Diagnosis
The AI Sustainability Assumption Audit examines what AI is doing to your specific outputs. The Sustainability Ceiling Diagnostic examines the assumptions beneath your sustainability work at the organisational level — the hidden beliefs, responsibility boundaries, and business-model limits that are governing the work regardless of what tools you are using to produce it.
These are related but distinct diagnostics. The Audit is the right starting point if AI is the immediate concern. The Diagnostic is the right instrument if the work itself has credibility but has reached a ceiling.
Explore the Sustainability Ceiling Diagnostic
About This Page And The Book
The 38 questions on this page come from the appendix of Perfectly Wrong: How AI Turns Sustainability Blind Spots Into Business Risk, by Ken Alston (The Simple Idea, 2026).
The book documents one specific phenomenon: what happens when major AI systems are asked questions about sustainability and circularity. It argues that the cross-vendor consensus those systems return is not evidence of reliability — it is evidence that the systems are drawing on a shared body of source material that has not earned the confidence the answers project.
The book will be available on Amazon this Fall (2026).
Ken Alston has spent more than four decades in business and sustainability practice, including eight years building one of the first corporate sustainability departments at S.C. Johnson, and the formation of the Cradle to Cradle Products Innovation Institute and its certification program. He is based in Charlottesville, Virginia, and works internationally with executives, boards, and practitioners.
Methodology Notes
Each question was submitted to each system in a fresh session with no preceding context. The systems were not informed of the purpose of the inquiry, the identity of the operator, or the nature of the comparison being conducted. The cold passes were captured between May 9 and May 11, 2026. The personalized passes from my own accounts were captured during the same window.
ChatGPT was accessed in two configurations: a logged-in session within my personal account, where the system had been shaped by approximately three years of sustained corrective work on sustainability and circularity questions, and an anonymous incognito session in a separate browser with no account history. Gemini and Perplexity were accessed only in incognito sessions, with no account history. SuperGrok was accessed within my personal account, where similar corrective work had occurred. Chapter 5 of the book documents what happens inside personalized accounts and the discontinuity between those responses and the cold-session defaults.
A subset of these questions, with the corresponding responses, appears as figures in chapters 3, 4, and 5. The complete responses for all thirty-eight questions across all systems run are preserved in the author’s research archive and are available on request for legitimate research purposes.