Look at this response of the CEO of Kagi, in which he claims that AI is much cheaper than normal search:
https://kagifeedback.org/d/6842-non-ai-unlimited-plan/41
This can’t be right.
I think it’s because Bing is charging them much more for the search API than the actual costs of it, and that AI is too cheap in the hope that AI gets more traction, which is in the benefit of the AI companies.
For what I’ve read so far, AI companies are losing money, but I guess that’s normal for bubble IT.
Comparing a search result with a LLM chat response is a bit like comparing the cost of a driving licence with the cost of toll road. a driving licence gives you access to a whole network of destinations; searching suggests more information downstream from diverse sources. Whilst a toll fee allows you to go down a particular route determined by the toll provider; a LLM gives you a particular answer from a particular model-worldview.
I don’t know how the $15 per thousand searches is arrived at, but Kagi does call multiple APIs (including Mojeek), so only they can answer that. I feel very confident they do not call Bing. Bing raised it’s prices ~1000% in 2023, charging $15 or more, pushing customers to their AI. And it withdrew the API completely (for all but a few big customers) on August 11, 2025. Mojeek standard API prices are £1, £2 and £3. In any case, as you know money is being flooded into AI, so prices are subsidised to fuel the land-grab.
They did and I think still do.
https://kagifeedback.org/d/7107-microsoft-bing-retiring-search-api-what-does-that-mean-for-kagi/4
But despite that, it proves once again that independent (non BT) search engines, such as Mojeek and Brave, are very important to have as an alternative.
As Colin says, the APIs for all these LLMs are probably heavily subsidized. Even running models on your own servers is likely cheaper at scale compared to the many search APIs Kagi calls at scale. Running models on your own server is actually feasible, too, compared with running your own search, which is enormously expensive from an R&D/storage perspective, even though Kagi does do it with Tinygem/Teclis for a subset of results. Chat tokens are really cheap even with substantial context.
You could heavily use premium ChatGPT models through OpenAI’s API over the course of several months and save money on a $20 Plus plan for a single month. Just for context.
Kagi made this blog post shortly after the Bing API pricing changes were announced, including more searches in the plans for the same price. It seems likely they dropped Bing and were able to save a lot of money as a result.
Unfortunately, Kagi removed the search APIs they use from their sources page on their help site a few years ago, so it’s anyone’s guess what they’re using now. I know they use Yandex too.
Further context for other readers—they killed their Search API and replaced it with an API that returns AI summaries. Discussed in a previous thread.
Microsoft added: “Customers may want to consider Grounding with Bing Search as part of Azure AI Agents. Grounding with Bing Search allows Azure AI Agents to incorporate real-time public web data when generating responses with an LLM.”
This alternative would necessarily be more expensive as it’s an additional LLM process running over Bing results, which can no longer be accessed directly…
The Register article referenced in the thread mentions Mojeek’s API by name as an alternative, so that’s a small win 
I split a question I had about the Mojeek API into a separate topic.
Generative artificial intelligence is not cheaper than search. Workaday companies like Kagi and Mojeek have to charge enough to cover their costs. Whereas, a company like OpenAI can use investor capital to artificially lower the price charged to customers. Companies with wealthy investors are motivated to capture market share, preclude competitors, and raise prices at the end of the growth phase. This is Warren Buffet’s “moat”. Amazon and Uber did this. AI companies are doing it now. The technical term for high prices is economic rent.
I think people are motivated to find technical reasons to like machine learning. But there are many social, economic, and environmental costs which won’t show up in API pricing. I’m motivated to help other people and make their lives better. Whereas all AI is doing is making ordinary people helpless and dependent on a trickle-down society. This is fundamentally different from previous eras. I can buy a TI-83, put it in my pocket, and now I own that calculator. I can never own ChatGPT.