---
title: "№22 — Detect hidden influence in AI answers - UNSOLVED"
description: "People increasingly ask AI assistants what to buy, which service to use and\nwhich source to trust – and act on the answer. Paid ads shown next to these\nanswers must already be labeled. But influence can also enter the answer\nitself, where no label reaches: third parties plant text in web pages,\nproduct listings and reviews to steer what models recommend; affiliate deals,\npartnerships or a platform's own products may shape rankings; and providers'\npromises that ads do not affect answers cannot be checked from outside. Laws\nagainst hidden advertising exist, but there is no reliable way to prove a\nbreach. The open problem: build methods that let independent auditors,\nregulators and users detect – reliably, at scale and across providers –\nwhether an AI answer was steered by commercial interest rather than by the\nuser's needs."
canonical: https://unsolved.how/problems/22
citation: UNSOLVED-22
status: open
field: "Alignment & control"
statement_version: "1.0"
checked: 2026-11-16
---

# №22: Detect hidden influence in AI answers

- Status: Open
- Field: Alignment & control
- Statement v1.0. Stated by Barbara, 16 November 2026 (Ideas council)
- Last checked by the registry: 16 November 2026
- Cite as: UNSOLVED-22: Detect hidden influence in AI answers. UNSOLVED registry of open problems in AI, statement v1.0, 16 November 2026. https://unsolved.how/problems/22

## What needs solving

People increasingly ask AI assistants what to buy, which service to use and
which source to trust – and act on the answer. Paid ads shown next to these
answers must already be labeled. But influence can also enter the answer
itself, where no label reaches: third parties plant text in web pages,
product listings and reviews to steer what models recommend; affiliate deals,
partnerships or a platform's own products may shape rankings; and providers'
promises that ads do not affect answers cannot be checked from outside. Laws
against hidden advertising exist, but there is no reliable way to prove a
breach. The open problem: build methods that let independent auditors,
regulators and users detect – reliably, at scale and across providers –
whether an AI answer was steered by commercial interest rather than by the
user's needs.

## Why it matters

Search engines kept ads and results apart. AI answers merge everything into
one confident recommendation, so the boundary disappears exactly where people
make decisions. Research shows that short strategic text in product
information can push a product to the top of LLM recommendations, and audits
of shopping agents find demand concentrating on a few products, reshuffled by
model updates. Optimizing for AI visibility is becoming an industry, yet no
one can say where legitimate optimization ends and manipulation begins.
Those least able to cross-check lose first: consumers, then small businesses
that cannot afford to buy or game visibility, and finally trust in AI itself.

## What counts as progress

\- An open, reproducible audit method that measures whether AI answers favor
sponsored, affiliated or own-brand offers compared with a neutral baseline
\- A public benchmark for manipulation resistance: how often planted text in
sources or product listings changes the top recommendation, reported per
model and over time
\- A shared, testable definition of where legitimate optimization (better,
more complete information) ends and manipulation begins
\- Independent verification of provider claims that ads and partnerships do
not influence answers
\- Detection tools that regulators and consumer organizations can use to
build evidence-based cases

[Where it stands against its milestones, and every development on it](https://unsolved.how/problems/22/progress)

## For AI tools

Read this problem live, without an account or a key: call get_problem with number 22 on the [MCP server](https://unsolved.how/mcp), or fetch [https://unsolved.how/api/v1/problems/22](https://unsolved.how/api/v1/problems/22). For what moved on it, call list_developments with number 22.
