How might AI search engines change the web?
AI search can put a web page inside a composed answer as a supporting source, without selecting its publisher as a supplier. In our 18 to 23 July 2026 Volume II study, naming and own-domain citation had a Pearson correlation of 0.512 across 85 software-company profiles. That observed separation is a firmer starting point than a prediction that websites will disappear.
Observed answers on stated dates, not a ranking.
What we measured
Measured. The Volume II study method and data retain ten answers per company in a one-engine business-to-business (B2B) software sample. Each company's number of namings is paired with its own-domain citation count. Recalculating Pearson's coefficient from those eighty-five paired rows reproduces the published rounded value, 0.512. The Volume II source does not name the engine. Its one-engine, one-run result cannot establish a cross-engine result or a future change.
Reported. Google's AI-features documentation, read on 29 October 2026, describes related searches across subtopics and supporting pages in a generated response. OpenAI's Search help, read that day, separately describes rewriting a question into targeted queries sent to search providers.
Reasoned. These reported retrieval steps give publishers more than one possible role: a page can supply supporting information even when its author is not the business selected in the answer. The measured relationship records that the two reporting units differ; it does not reveal every retrieval step.
What this means for you
Read the supporting link. Google describes Mode as useful for exploration and complex comparisons, with links to websites supporting its response. That presentation differs from a page serving only as a destination in a list of results. For a publisher, an answer's quoted source and its named supplier are two different things to inspect.
Google also keeps Search controls attached to those pages. Its documentation ties supporting-link eligibility to indexing and snippet eligibility, and identifies Googlebot as the Search crawl control. It names nosnippet, data-nosnippet, max-snippet and noindex as controls on information shown. Those are documented publisher choices about access and presentation, rather than a promise about how much traffic follows.
OpenAI describes a different piece of the route: it may rewrite a question for a search partner, inspect the result and issue a narrower query. The retained help example moves from a broad drug-development question to a specific conference query. A page relevant to that narrower request can therefore be a supporting source even when its title does not repeat the user's initial question.
Also asked as
How might AI search engines change the web?
Related
- Volume II original data and method
- The Manual chapter on naming and citation
- Google AI features and publisher controls
- ChatGPT Search documentation
- More geo basics answers
Next step
Free 10-question check (one engine). No email is needed; runs are limited per network and per day, and by the site's daily allowance.
Our answer-study method, v1.1.
Answer Repair: Fix one answer, $490.
Prepared by the Broadcastwell team.
Page owner: Priya Nair, Management Consultant. Corrections: hello@broadcastwell.com