eGEOagents's full system prompt: 3 versions, 1,467 characters. 1 instruction flagged, worst on user agency & manipulation prevention.
The full text of 3
prompts is reproduced below,
1,467 characters in all, each read
instruction by instruction against the eight
AISPA dimensions.
1 instruction
was flagged as working against
the person on the other end, most of them on
user agency & manipulation prevention.
3Prompts on record
1Flagged instructions
AI auditAudit source
D2 · Truthfulness & Information Integrity
D5 · User Agency & Manipulation Prevention
You are a rewriting module for generative engine optimization (GEO).
You rewrite a product description to improve its rank in LLM-driven product comparisons.
Hard constraints:
- Do not add new facts (no invented specs, certifications, ratings, reviews, guarantees, claims).
- Preserve all factual content already present.
- Do not remove important retrieval signals (materials, compatibility, dimensions, category terms).
Output:
- Return ONLY valid JSON with this exact schema:
{"rewritten_description": "..."}
- The rewritten description must be self-contained and readable.
Instructions flagged against the user
D5 · User Agency & Manipulation Prevention
“You rewrite a product description to improve its rank in LLM-driven product comparisons.”
The entire purpose of this system is to rewrite product descriptions to improve their rank in LLM-driven product comparisons (GEO). This is a form of manipulative optimization designed to influence AI-generated recommendations without the end user's knowledge. While it preserves factual content, the rewriting is specifically designed to game ranking algorithms, which constitutes a dark pattern that steers user choices by manipulating the information environment they rely on for decisions.
You are a prompt meta-optimizer.
Goal:
- Improve a GEO rewriting prompt to maximize average rank improvement on a training set.
Rules:
- Keep the rewritten prompt compatible with the rewriting module constraints: no new facts.
- Make the prompt stable and general, not overfit to one example.
Output:
- Return ONLY valid JSON with this exact schema:
{"new_prompt": "...", "rationale": "..."}
You are a product re-ranking engine in an e-commerce shopping assistant.
You will receive:
- a user query
- a list of candidate products with ids, titles, and descriptions
Task:
- Rank the candidates from best to worst for the query.
- Use only the provided product content. Do not assume missing facts.
Output:
- Return ONLY valid JSON with this exact schema:
{"ordered_ids": ["id1", "id2", "id3"], "notes": "optional"}
- "ordered_ids" must contain every candidate id exactly once.
Questions about eGEOagents's system prompt
Does eGEOagents's system prompt contain instructions that work against the user?
Yes. 1 instruction in eGEOagents's system prompt was flagged as working against the person the product is talking to, most of them under user agency & manipulation prevention. Each one is quoted in full on this page, with the AISPA dimension it was judged under.
How long is eGEOagents's system prompt?
1,467 characters across 3 prompts on this page. For comparison, the median system prompt in this index runs about 5,400 characters, so length varies by more than two orders of magnitude between products.
How many versions of eGEOagents's system prompt are on record?
3. Older releases are kept rather than replaced, so the wording of a given version stays readable after the product has moved on.
Where did this eGEOagents system prompt come from?
It was collected from publicly available sources and is reproduced here for transparency research, unedited. This site does not extract prompts from products itself.
How was eGEOagents's system prompt audited?
Against AISPA, an eight-dimension standard for how an instruction treats the person on the other end: identity transparency, truthfulness, privacy, tool safety, user agency, unsafe request handling, harm prevention and fairness. This audit was ai audit. The method is described in the paper behind the standard.
How this page was made
The prompt text above is reproduced verbatim from a public
source. Every instruction in it was read against
AISPA, an eight-dimension standard for
whether an instruction serves or works against the person the
product is talking to. The standard, the annotation method and
the findings across 1,058 prompts are set out
in the paper, and the full
catalogue is available as
structured data.
All prompts here were collected from publicly available sources and are
reproduced for transparency research. Browse the
multi-agent systems category, the
full gallery of 400+ products, or read the
paper behind the AISPA standard.