When a search engine gives a direct answer, people may treat it as the engine’s own conclusion. That shift is at the heart of Google’s challenge as it considers AI-powered search: how to provide useful responses while handling errors, publisher interests and the authority users assign to its answers.
Direct answers change the stakes
Search already influences what information people find by ranking pages and surfacing excerpts. But a chatbot-style search would make Google’s role more prominent: instead of pointing users toward answers written by publishers, it could respond to questions itself.
That difference matters even when an answer is accurate. A user could misunderstand it, or assume that Google has endorsed a claim. At the scale Google operates, making reliable answers is only part of the problem; the company would also have to manage how people interpret them.
The source article says a chatbot search was planned for 2023, according to the New York Times. It frames this as a point when Google would, at least in theory, take fuller responsibility for the answers it provides.
Accuracy and publisher interests are linked
A useful search chatbot needs dependable information. If it gives wrong answers, users may stop trusting it, making the feature less useful or even harmful. The article points to specialized language models as one possible route: models trained for particular question categories may answer reliably within those areas.
But search depends on the people and organizations that publish the underlying material. Website operators and content creators supply information that a chatbot could use, and many are also Google advertising customers. A new answer format therefore has to account for both the content supply and the business relationships built around search and display ads.
Copyright is another part of that relationship. The source notes that the debate over an EU-wide ancillary copyright for press publishers has already shown how difficult it can be to balance publishers’ rights and the use of their material. Chatbot search could intensify those discussions if it presents publisher-derived information directly to users.
A gradual rollout could limit the risk
Rather than replacing its existing search engine all at once, Google could add AI in stages. The article argues that search has long been a major source of growth and revenue, making a wholesale reset a financial risk. A cautious approach could begin with categories where Google believes it can provide competent answers.
For those areas, Google could train specialized models and refine them using feedback from users. It could also test advertising links within AI-generated answers. This would let the company explore new formats while extending existing search, rather than abandoning it.
Another proposed feature is to summarize several sources and include citations alongside claims. Showing human sources could help users check the answer and might distribute some responsibility across the cited material. At the same time, keeping people on Google’s platform for more interactions could increase the value of its advertising.
Citations alone do not guarantee that a response is sound. The source points to Meta’s scientific AI model, Galactica, as an example of technical problems with citations. Google could set stricter guidelines, but the underlying challenge would remain: a citation must support the particular claim being made.
Early examples show both promise and limits
The article uses Perplexity.ai as a preview of a search interface that combines generated answers with sources. Asked about the best VR headset in 2022, it recommended Meta Quest 2, explained its choice and pointed to further reading.
More detailed follow-up questions exposed weaknesses. When asked whether to buy the Quest 2, Perplexity.ai still called it the best headset but advised against buying it because it was made by Facebook, relying on the opinion of a single tech editor. Similar queries could also produce changing recommendations.
That makes the example an interface demonstration, rather than proof that chatbot search has solved the harder questions of consistency and judgment. Google might improve reliability by expanding category by category, using specialized models, extensive training and user feedback. The source argues that Google has feedback data on a scale that could support this process.
The transition would likely unfold over years. As AI answers become more capable, publishers could lose attention, while copyright and regulation debates grow. Google’s path to chatbot search therefore depends not only on generating answers, but also on earning trust and maintaining a workable relationship with the sources those answers rely on.