ChatGPT Search

ChatGPT Search is OpenAI’s web-search feature integrated into ChatGPT. It retrieves live web results, reranks them for transparency, and emits responses with inline citations. Bing index ranking and Wikipedia presence are the strongest signals.

How ChatGPT Search works

ChatGPT Search is activated when ChatGPT determines that a query requires current or factual information not reliably available in its parametric knowledge. It issues queries to Bing (and OpenAI’s own crawler for certain content types), retrieves candidate pages, reranks them using a combination of relevance and credibility signals, and synthesises a response with inline citations. The citations appear as numbered superscripts linked to the source pages.

The system is distinct from the base ChatGPT model: parametric knowledge governs responses when Search is off; retrieved documents govern responses when Search is on. A brand’s visibility in the two modes can differ substantially, particularly for European brands whose English-language web presence is thinner than their domestic reputation would suggest.

Domain authority proxies in ChatGPT Search

Research characterising retrieval-augmented LLMs selecting from web search results (arXiv:2506.00054) found that domains with high topical authority — specialist publications, government sources, academic institutions — are preferentially cited even when semantically equivalent content is available from lower-authority domains. ChatGPT Search applies a version of this: Bing ranking, domain age, inbound link profile, and Wikipedia entity presence all function as authority proxies.

For European brands, this creates a compounding disadvantage: weaker English-language web presence means weaker Bing ranking, which means lower ChatGPT Search citation probability. The gap is not overcome by publishing volume alone — it requires earning independent editorial coverage on domains that Bing already ranks highly.

Comparison with Perplexity

Perplexity operates a similar web-grounded RAG architecture but uses its own crawler and a broader index that includes academic preprints (arXiv, PubMed). Perplexity tends to return shorter responses with more citations per response than ChatGPT Search. In CEAVERS Q2 2026 data, Perplexity yielded the highest mean brand visibility score across the European panel (mean +3.7 over the six-LLM mean) — likely due to its higher citation density and inclusion of academic sources.

robots.txt configuration

Allow GPTBot and OAI-SearchBot explicitly in robots.txt. Both are required: GPTBot governs training data crawls; OAI-SearchBot governs Search retrieval. Blocking either reduces ChatGPT Search visibility.