Quick Overview
- ChatGPT (GPT-4 and later versions) was trained on web data up to various cutoff dates.
- This is where effective action is most achievable.
- Every AI retrieval system - whether it's Perplexity pulling live pages or ChatGPT with browsing enabled - depends on indexed web pages.
You've addressed your Google results. But now someone asks ChatGPT about you - and it gives them a summary of your court case. AI search is a new frontier for online court records. Here's what's actually happening inside these systems, and what you can realistically do about it right now.
How AI Chatbots Surface Court Records
Understanding why AI systems know about your court record requires understanding the difference between two fundamentally different types of AI systems:
Type 1: Large Language Models With Training Cutoffs (ChatGPT Base, Gemini Base)
Models like ChatGPT (non-browsing version) and Google Gemini (without real-time search) were trained on massive web crawls that happened to include court record websites. Sites like Justia, CourtListener, PACER, and state court portals were crawled as part of training data collection. If your record appeared on any of those pages during the crawl window, that information may be encoded within the model's parameters - the billions of numerical weights that define its behavior.
This is fundamentally different from a database. The information isn't stored as a file that can simply be deleted. It's diffused throughout the model's weights in ways that a standard deletion request does not target, which is why the work centers on the source pages.
Type 2: AI Search Engines With Real-Time Retrieval
AI search engines like Perplexity AI, Google AI Overviews, and Bing Copilot work differently - they retrieve live web pages at the time of your query and synthesize the content. If your court record page is currently indexed on the web, these systems can surface it right now, in any query.
ChatGPT and Trained Data
ChatGPT (GPT-4 and later versions) was trained on web data up to various cutoff dates. OpenAI has released several model versions with different training cutoffs. If court record websites were indexed during the training crawl, the model may have absorbed that information.
How to Check If ChatGPT Knows About Your Record
- Open a ChatGPT session (free or paid).
- Ask directly: "What do you know about [your full name] from [your city/state]?"
- Follow up with variations: "Have you seen any legal or court information about [your name]?"
- If ChatGPT surfaces your record, note what information it provides - this helps identify which source documents were included in training data.
OpenAI's Deletion Request Process
OpenAI maintains a privacy portal and processes requests under applicable privacy laws. Options include:
- Subject Access Request (SAR): Request to know what personal data OpenAI holds about you.
- Deletion request: Request deletion of personal data under CCPA (California residents) or GDPR (EU residents). Contact: privacy@openai.com or use their privacy portal.
- Opt-out of training: You can request your conversation data not be used for training via account settings.
Requesting deletion from OpenAI's training data tends to be less effective for web-sourced content than removing the sources. OpenAI cannot surgically remove specific facts about a person from model weights without retraining - and even then, the same information may be relearned from other sources. The more reliable path: remove the source web pages so that future model versions don't have access to them.
AI Search Engines With Real-Time Access
This is where effective action is most achievable. AI search systems that retrieve live web content are entirely dependent on what's currently indexed and accessible.
Follows Google's De-indexing Rules
If a URL is de-indexed from Google (via removal tool or robots.txt), Google AI Overviews cannot surface it. De-indexing from Google = removed from AI Overviews. This is the most actionable solution for Google's AI system.
Real-Time Web Retrieval
Perplexity retrieves live web content. There is no formal removal process with Perplexity - but removing or de-indexing the source pages prevents Perplexity from surfacing them. Source removal is the solution.
Follows Bing's Removal Policies
Bing Copilot uses Bing's index. Submit removal requests through Bing's Content Removal tool for outdated or sensitive content. Content removed from Bing's index is removed from Copilot responses.
Real-Time + Trained Data
When ChatGPT uses its browsing capability, it can visit and summarize live web pages. If your court record page is live and indexed, ChatGPT with browsing can retrieve it. Remove the source → browsing mode can no longer access it.
How to Request Removal From AI Companies
Each major AI company maintains privacy processes, though effectiveness varies:
- OpenAI (ChatGPT): Privacy portal at privacy.openai.com. Submit deletion requests under CCPA or GDPR. For business accounts, contact privacy@openai.com.
- Google (Gemini, AI Overviews): Use Google's existing personal information removal tool for content that violates their policies. De-indexing through Google Search Console or URL removal tool applies to AI Overviews automatically.
- Microsoft (Bing Copilot): Use Bing's Content Removal tool. Requests for outdated cached content and personal sensitive data are processed separately.
- Perplexity: No formal public removal process as of mid-2026. Contact via their support channels. Source removal is the most effective approach.
- Anthropic (Claude): Privacy requests via privacy@anthropic.com. Follows similar principles to other major LLM providers on data removal.
The Most Effective Strategy: Remove the Source
Every AI retrieval system - whether it's Perplexity pulling live pages or ChatGPT with browsing enabled - depends on indexed web pages. The single most effective strategy for AI record removal is eliminating or de-indexing the source pages that AI systems read from.
This means systematically addressing:
- Justia and CourtListener: Direct removal requests; documentation of sealing or expungement accelerates approval.
- State court portals: Varies by state - some allow redaction petitions; others require court orders.
- Background check and data broker sites: Opt-outs for Spokeo, BeenVerified, Whitepages, Intelius, PeopleFinder, and others.
- Google de-indexing: Removes source pages from Google's index - and correspondingly from Google AI Overviews.
- Bing removal: Parallel process for Bing's index, which feeds Copilot.
- Mugshot sites: Arrest photo pages that AI systems may read and summarize.
State-by-State AI and Privacy Laws
The legal landscape for AI-related privacy rights is evolving rapidly:
- California (CCPA/CPRA): Provides the strongest privacy rights currently available in the US. California residents can request deletion of personal data from AI companies and opt out of the sale or sharing of personal information. The California Privacy Protection Agency (CPPA) is actively developing AI-specific regulations. See California's CCPA overview.
- Colorado, Virginia, Connecticut, Utah: Have passed comprehensive privacy laws with data deletion rights applicable to AI systems that process personal data.
- EU GDPR: The most comprehensive AI data rights globally - includes the "right to erasure" (Article 17) and the right not to be subject to purely automated decisions with legal effects (Article 22). EU residents have the strongest practical tools for AI data requests.
- Federal law (US): The FTC has been active on AI data practices under its Section 5 authority. No comprehensive federal AI privacy law has passed as of mid-2026, though several proposals are moving through Congress.
"Transparent, specific, and honest from day one. They told me which records were removable and which weren't before I paid anything."
Is your court record appearing in ChatGPT? Find out - free.
Tell us about your situation and a removal specialist will personally review it and respond within one business day. No pressure, no obligation.
- No upfront payment - you only pay if we succeed
- A+ BBB Rated · 5,000+ Cases Handled · Since 2013
- 100% Confidential · Response within 1 business day
