OpenAI

ChatGPT

Optimise for the 200M weekly user surface

ChatGPT is two engines wearing one wrapper: a language-model recall path and a live-browsing path. Visibility requires winning on both.

Quick facts

Vendor

OpenAI

Weekly users

200M+

Crawlers

GPTBot, OAI-SearchBot

Sourcing

Hybrid (training + browse)

Citation pattern

Browse mode only

Language coverage

All major

How ChatGPT decides what to say about your brand

By default ChatGPT answers from training data — a snapshot of the public web frozen at the model's training cutoff (currently around early 2025 for the latest model variants). When the question contains time markers like „latest" or „2026", or when the model is uncertain, it triggers the live browse tool, which fetches real-time web pages and feeds them into the answer.

Visibility on the training-data path depends on **entity prevalence** — how often your brand co-occurs with the topic across the training corpus. Wikipedia stubs, Crunchbase entries, news mentions and structured directory presence are the biggest single-pixel improvements.

Visibility on the browsing path depends on **search-backend ranking + first-fetch parseability**. If your page is JS-heavy, behind auth, or returns slow, the browse tool will skip it.

What you can do this week

Audit robots.txt — allow GPTBot AND OAI-SearchBot. Verify with curl.
Check `<head>`: ship Article + FAQPage + Organization JSON-LD on cornerstone pages.
Confirm server-side rendering of H1, lead paragraph and key facts.
Add `dateModified` to all evergreen pages and update them substantively.
If you're not on Wikipedia, identify two journalists or industry-list editors to pitch a stub.

How to measure ChatGPT visibility

Manual: pick five buyer-relevant prompts, run them in ChatGPT once with browsing on, once with browsing off. Note whether your brand appears, how, and which sources are cited (when browse is on).

At scale: a daily prompt-replay tool — like GEOnative — runs your prompt set across ChatGPT (both modes) and produces longitudinal Visibility, Share of Voice and citation data. Manual tracking does not scale past ten prompts.

How GEOnative helps

GEOnative tracks ChatGPT visibility across both training-data and browsing modes daily. We capture which sources are cited, classify sentiment of every mention, and alert you when your visibility shifts week-over-week.

Try GEOnative →

FAQ

Does ChatGPT show citations?

Only when browsing is enabled. Default-mode ChatGPT (training-data path) gives prose answers without source links. ChatGPT Search and the browse tool inside the standard chat surface both show clickable citations.

Should I block GPTBot?

For most brands, no. Blocking GPTBot excludes you from one of the largest AI surfaces. The few legitimate reasons to block — paid-only content, regulated information, IP concerns — apply to specific pages, not whole domains. Block at path level, not site level.

How fast does ChatGPT reflect new content?

Browsing mode picks up new content within hours of indexing. Training-data mode lags 6–12 months — the time between training-data cutoffs. New content is best for browsing visibility; entity-presence work for training-data visibility.