A score dip is concerning. A complete disappearance is a business problem. If your situation is a gradual drop rather than a sudden absence, the causes of a falling AI visibility score covers that pattern separately.
If your brand was reliably appearing in ChatGPT responses three months ago and is now absent across tested queries, something specific changed. This guide walks through the five most common causes of complete citation loss and what to do about each one in the first 72 hours.
This covers complete disappearance — your brand stops appearing across multiple query types simultaneously. If your visibility score has dipped but you still appear in some queries, that is a different pattern with different causes.
Why Complete Disappearance Happens
ChatGPT and other large language models do not constantly re-crawl the web. Their primary training data has a cutoff, and their real-time retrieval (when active) depends on web crawlers that follow the same rules as search engine bots. That means certain technical failures can cause a brand to effectively vanish from the information landscape that AI models draw on.
Complete disappearance almost always has a traceable cause. Here are the five patterns that account for most cases.
Cause 1: Robots.txt Now Blocks AI Crawlers
A change to your robots.txt file — whether intentional or made by a developer as a side effect of another update — may have blocked GPTBot, ClaudeBot, PerplexityBot or other AI crawlers. These bots check robots.txt before crawling and will not index pages they are denied access to.
Go to yourdomain.com/robots.txt. Look for Disallow rules that apply to User-agent: * or to any of the named AI crawlers. A common accidental block looks like: User-agent: * / Disallow: / added during a site migration or CMS update.
Cause 2: Knowledge Graph Entity Detachment
AI models rely on entity graphs to understand what a brand is, what it does, and how it relates to other entities. If your brand was connected to relevant nodes — product categories, founder information, industry associations — and those connections became contradictory or were removed, the model may no longer be able to reliably place your brand in a recommendation context.
This typically happens when:
- Your Wikidata entry was edited to remove or contradict key facts
- Your company description changed on major reference sites (Crunchbase, LinkedIn company page, Wikipedia) without being updated consistently
- A product rebrand created conflicting entity names across different sources
Search your brand name on Wikidata. Check that your company description on LinkedIn, Crunchbase, and your homepage Organisation schema all say the same thing about what you do and what category you are in. Contradictions create entropy that can remove you from reliable recommendation pools.
Cause 3: Negative Third-Party Signal Spike
A sudden increase in negative content about your brand on high-authority platforms — Reddit, G2, Trustpilot, major industry publications — can shift how AI models characterise your brand. If the dominant signals about your brand switched from neutral-to-positive to predominantly negative, models that prioritise user-consensus sources may exclude you from recommendation responses to avoid recommending something with a contested reputation.
Search your brand name on Reddit filtered to the last three months. Check G2 and Trustpilot for recent review spikes. If you find a cluster of negative content that appeared suddenly, that is likely a contributing factor.
Cause 4: A Competitor Updated Their Schema or Entity Coverage
Your disappearance may not have been caused by anything you did. If a competitor added comprehensive Organisation schema, earned a Wikipedia entry, or secured coverage in several authoritative publications simultaneously, they may have become the dominant entity for queries where you previously appeared. AI platforms do not have infinite space in a recommendation response — stronger entity signals displace weaker ones.
Run your standard test queries and document which brands are appearing where you used to. Check their websites for recent schema additions (view source and search for "@type"), their Wikipedia or Wikidata pages for recent edits, and their PR coverage from the last 60 days.
Cause 5: Model Update or Training Cutoff Shift
ChatGPT, Gemini and other platforms update their underlying models periodically. A new model version may have different training data, different grounding priorities, or different retrieval patterns. Brands that appeared consistently in an older model version may need to build fresh signals to appear in a newer one, particularly if they are in a competitive category where many options exist.
Note when you first observed the disappearance. Cross-reference with public announcements about ChatGPT or Gemini model updates. If the timing aligns, the solution is to build new, authoritative signals rather than trying to restore the previous state — the previous model is no longer serving responses.
Your 72-Hour Triage Protocol
Once you have identified the cause and made your fixes, consider setting up systematic monitoring so you catch future drops before they become full disappearances.
- Check robots.txt for accidental AI crawler blocks
- Verify Organisation schema is present and correct on your homepage
- Audit entity consistency across Wikidata, LinkedIn, Crunchbase and your homepage
- Search Reddit, G2 and Trustpilot for negative content spikes in the last 90 days
- Run ten test queries and document who replaced you
- Check competitor websites for recent schema or Wikipedia additions
- Note whether the disappearance timing aligns with a known model update