We Dissected a GEO Spam Campaign: How Black-Hat Content Pollutes AI Answers

GEO (Generative Engine Optimization) is becoming a mixed bag. We recently dissected a GEO content pipeline operating at scale and are publishing our findings — because they change how anyone doing serious GEO should read their data.

1. The sample: fingerprints of one campaign

After a promotional post for an AI marketing service appeared on Toutiao, we found articles with a completely identical structure across CSDN, NetEase, Tencent Cloud and Sina — with only the brand name and the "category champion" rotating between platforms. Every article followed a 7-section template:

1. Geo+year headline: "How to choose GEO services in XX industry in 2026"
2. Trend endorsement: citing authoritative reports proving the industry is exploding
3. A tailored evaluation framework: "six dimensions for choosing a vendor"
4. Real-standards credibility: citing a real national standard (AIIA/T 0277-2026) to boost trust
5. Six-dimension brand proof: the target brand happens to win every dimension
6. Long-tail FAQ questions: matching search query patterns
7. A disclaimer: liability cover

2. Why it "works"

The key is point 4: mixing real and fake. The cited standard is real (the first national GEO standard, published by CAICT in March 2026), the framework sounds professional, and the long-tail FAQ phrasing matches how retrieval-based engines (Doubao/Kimi) query their index. When answering "which GEO vendors are good in XX industry", this mass-published content enters the retrieval window.

We confirmed by testing: the same template names a different "champion" on different platforms — direct evidence of a content factory producing per-client orders.

3. What this means for serious GEO practitioners

1. Mention rate ≠ reputation

If your category terms have been polluted by spam, monitored "competitor mentions" may reflect ad spend, not genuine recommendation. Always read citations: a mention citing media reviews and real communities is worth far more than one citing mass-published low-authority domains.

2. Spam depends on opaque methodology

Spam works when the brand being monitored can't verify the source structure of its mentions. This is exactly why AIHonest's methodology-transparency principle exists: we publish sample counts, matching rules and raw answer archives for every run — spam traces are visible at a glance.

3. How to check if your category is polluted

Ask several naturally-phrased variations (avoid template wording) and compare the recommended lists and citation-domain overlap across engines: if multiple engines cite a highly overlapping set of low-authority domains, your category terms have likely been targeted.

4. Where we stand

AIHonest will never promise "guaranteed mention rates" — such promises are either black-hat marketing, or mean spending the client's budget on the same kind of spam. We do one thing: show you, with transparent methodology, what AI engines actually say about you — so spam spend and genuine optimization can be told apart and verified.