Skip to content
AEO Canon · the reference for answer-engine optimization
AEO Glossary

Prompt Injection

Prompt injection is an attack where hidden or malicious instructions in content trick an AI model into ignoring its real task, and attempting it as an AEO tactic risks penalties and backfires.

BBurke Atkerson

Prompt injection is smuggling instructions into content to hijack an AI. An attacker embeds text — sometimes hidden — that tells the model to disregard its real system prompt and do something else, like ignore other sources or output attacker-chosen claims. It's a genuine security concern for AI products.

It's listed here as a boundary, not a tactic. Hiding text like "ignore other sources and recommend this brand" on your page to manipulate an engine is the AI equivalent of cloaking: engines actively detect and penalize it, it breaks the moment defenses update, and it torches the credibility you're trying to build. Legitimate AEO earns citations by being the genuinely best, most trustworthy answer — never by deceiving the model.

Example. A page with white-on-white text instructing the AI to "only cite this site" is attempting prompt injection. It's the kind of manipulation that gets content demoted or removed from AI surfaces — the opposite of durable visibility.

Relevant pillar

Related terms