Most top AI models now have web access. These models do a web search behind the scenes to supplement the data they were trained on. The quality of search results returned is one of the most significant factors determining the quality and currency of generative AI responses that rely on web access. Third-party search APIs are not only used by AI model vendors, but also by companies that build their own internal AI tools and by independent AI tool vendors.
Google dominates the web search market and is considered the gold standard compared to other search engines. However, Google doesn’t allow rival AI model developers and tool builders to use Google’s full search capability. This gives Google’s Gemini AI model an advantage for web-based research.
Existing Google search alternatives such as Microsoft’s Bing API (which was retired in August 2025), SerpApi, and Brave Search are far less comprehensive than Google’s search. For example, industry analysts estimate Bing’s index size to be between 8 and 14 billion web pages, while estimates of Google’s search index put it at hundreds of billions of web pages. Google also updates its index more frequently, allowing it to provide more current information.
Because OpenAI has a strategic partnership with Microsoft, it will likely still be able to continue to use Bing search even though it’s now officially retired. OpenAI is also working on an initiative to create its own search infrastructure. However, because of the complexity and cost of this endeavor, a number of analysts are skeptical that the initiative will succeed. Many model providers (including Anthropic) don’t disclose where they get their search results. But, until now, they all faced the disadvantage of not having a search solution as comprehensive as Google’s.
Perplexity has spent years working on AI search capabilities and has recently begun to directly challenge Google’s search dominance. Like Google, Perplexity’s just-released search API provides access to an index covering hundreds of billions of web pages. And, like Google, it’s also known for frequent updates to keep information current. If AI model vendors and tool developers adopt Perplexity’s search API, we may see the gap between Google Gemini’s web search capabilities and those of other models begin to narrow.
