Is Your Brand Being Poisoned? Competitor Analysis for AI Visibility

A prospective customer asks an AI assistant to compare your product to a competitor's. The answer it gets back was shaped, in part, by what that competitor chose to publish, how they structured their data, and — less obviously — how they've configured which AI crawlers can even read their site. None of that is under your control, but it's worth understanding, because it's actively shaping how your brand shows up in AI-generated answers whether you're watching or not.

What "AI poisoning" means here (and what it doesn't)

Used loosely in marketing and SEO conversations, "poisoning" in this context refers to competitors shaping an AI system's perception of your brand through legitimate but one-sided publishing — comparison pages, "alternatives to [you]" content, review roundups — not to any technical attack on a model's training pipeline. That's an important distinction: nothing here is illicit or requires a security response. It's closer to old-fashioned competitive positioning, just operating on a new surface where the audience is an AI system synthesizing an answer rather than a human reading a page directly.

The practical effect is the same either way. If a competitor has published five well-structured "X vs. [you]" comparison pages and you've published none, an AI system asked to compare the two has five times as much competitor-authored material to draw from about your own product as you've given it yourself.

What to actually check on a competitor's site

A few concrete things reveal how a competitor is positioned for AI visibility, and by extension how they're likely showing up in comparison queries:

• Comparison and "alternative to" content. Search a competitor's site for pages structured around your brand name — these are usually written to be cited, with clear head-to-head claims that an AI system can quote directly.
• robots.txt. Check whether a competitor blocks or allows AI crawlers like GPTBot, ClaudeBot, and PerplexityBot — a blocked competitor is invisible to those crawlers regardless of what they publish, which changes where the AI system's information about them (and about the comparison) is coming from.
• llms.txt. If present, it's a direct index of what a competitor wants an AI model to treat as their key pages — useful context for understanding what they're actively steering models toward.
• JSON-LD structured data, particularly Product, Offer, and Organization schema — these are the fields an AI system is most likely to extract as fact rather than paraphrase, so a competitor with complete, accurate schema has a structural advantage independent of the prose on the page. See which schema types matter most for AI for the specific fields worth checking.

None of this requires special tools — a browser, view-source, and a curl request to /robots.txt and /llms.txt cover the basics.

It's also worth checking how a competitor's comparison content is structured, not just that it exists. A page that states specific, structured claims — a table of features, a schema-marked list of differentiators — is more likely to be extracted and repeated by an AI system than a page making the same points only in loosely organized prose. Two competitors can publish functionally the same comparison, and the one with cleaner structure is the one more likely to actually shape what a model says.

Monitoring strategies

Ad hoc checking works as a one-time audit but goes stale — a competitor can publish a new comparison page or change their structured data at any time, and there's no notification when it happens.

Recurring, systematic monitoring is what actually catches change over time. CrawlPod's Competitor Poisoning Alerts automates this: it scans defined competitor sites on a schedule, checks for new or changed comparison content, crawler configuration, and structured data, classifies the resulting threat level, and sends an email alert when something material changes — rather than requiring a manual re-check to notice.

Turning monitoring into a measurable position

Knowing what a competitor is doing is only half the picture; the other half is knowing where your own brand actually stands in AI-generated answers relative to them. CrawlPod's AI Share of Voice Report measures how often, how favorably, and how accurately your brand is mentioned across AI answer engines compared to named competitors — the outcome-level complement to watching what competitors are publishing on the input side.

What a threat classification actually tells you

Not every change on a competitor's site is equally significant. A minor copy edit to an existing comparison page is a different situation than a competitor newly blocking your brand's structured data from being referenced favorably, or publishing a comparison page with a materially misleading claim about your product. Distinguishing low-significance noise from a genuinely material change is what a threat-level classification is for — it's the difference between an alert worth acting on immediately and one worth simply logging for context, and it's the part of monitoring that's hardest to do consistently by hand over time, since judgment about "does this actually matter" tends to drift depending on how recently a reviewer last looked closely.

A reasonable response, not an alarmed one

Discovering that a competitor has ten comparison pages and you have zero isn't evidence of anything nefarious — it's evidence of a content gap, the same kind that's existed in SEO for two decades, now showing up on a new surface. The response is proportionate: publish accurate comparison and alternatives content of your own, keep structured data complete and current, and make sure AI crawlers can actually reach your site in the first place — see the WordPress AI visibility guide if that's the platform in question. Ongoing awareness of what competitors are doing is useful context for deciding where to invest that effort next, not a reason to treat every competitor publish as a threat.