Google, OpenAI and McKinsey numbers point to the same shift: brands now need to be cited inside AI answers, not just found in search results. For most communications teams, that gap is already dangerous.
For years, corporate communications teams had a comfortable assumption: publish the announcement, get the coverage, build the backlink, wait for Google to index it.
That system is not dead. But it is no longer enough.
The first stage of research is moving away from a page of search results and into AI-generated answers. A buyer no longer has to search “best enterprise software for compliance”, open five tabs and compare supplier websites. They can ask ChatGPT, Claude, Perplexity, Gemini or Google AI Overviews for a shortlist.
That creates a new problem for brands. The press release may exist. The website may be live. The SEO strategy may be respectable. But if an AI system cannot understand the claim, trust the source or cite the evidence, the brand may be absent from the answer.
This is why AI citations are becoming a serious marketing and reputation issue. Not because companies can pay their way into ChatGPT, but because they need to make their facts easier for machines and humans to verify.
The numbers are now difficult to dismiss. At Google I/O 2026, Sundar Pichai, chief executive of Google and Alphabet, said AI Mode was “our biggest upgrade to Search ever”. The audience figures tell their own story:
That last figure is the one many marketing teams have still not absorbed. AI is not just another publishing channel. It is becoming an advisory layer.
If an AI system cannot understand the claim, trust the source or cite the evidence, the brand may be absent from the answer.
If a potential client asks an AI assistant which law firms understand fintech regulation, which consultancy is strong in AI transformation, which cybersecurity vendor is credible on ransomware — the answer may shape the shortlist before anyone reaches a website.
McKinsey has called AI-powered search the “new front door to the internet”. Its analysis puts concrete numbers on the shift:
For boards, this should be uncomfortable. The old visibility model assumed that traffic came first and persuasion came second.
AI search reverses that. The persuasion can happen before the click.
That makes much of today’s corporate content look weak. A company says it is “a leading provider”. It announces a “strategic partnership”. It claims to deliver “transformational outcomes”. These phrases may be normal in a press release, but they are poor evidence.
They do not say much. They are hard to verify. They are not distinctive. Most importantly, they give AI systems very little to cite.
A better claim is precise. It includes a number, a date, a named person, a named organisation or a source. Consider the difference:
The first statement is marketing language. The second is a citation.
Traditional SEO is still important. Strong pages, clear structure, technical performance, authoritative backlinks and useful content all matter. But AI visibility adds another layer. The question is not only whether a page can rank — the question is whether a fact can be selected as evidence.
Recent research suggests those are not the same thing. A 2026 study of Google AI Overviews analysed 55,393 trending queries over a 40-day period and found:
That third point should concern any company relying only on classic SEO reporting. A brand can rank and still be ignored. Another source can be cited even if it is not one of the traditional blue links.
A 2026 study of LLM brand reputation went further, analysing 167,551 URL-grounded citations across 128 brands, 12 home markets and 13 languages:
Press coverage is no longer just about human readers. It may also become part of the machine-readable reputation file.
That finding should change how companies think about PR. The brand’s own website matters, but it is not the whole evidence layer. What publishers, analysts, directories, review platforms, trade bodies and specialist sites say about the brand may be more influential than the brand’s own claims.
That does not mean every paid article or press release has value. Thin promotional content is still thin. The future belongs to content that can be used. Specifically, content that includes:
This is where many PR agencies are behind the curve. They still optimise for publication rather than citation. They care whether the article goes live, whether it has a link, and whether the client name appears. Those things still matter, but they are no longer sufficient.
The better question is: what fact does this article add to the public record?
For City firms, this matters across sectors. Compliance leaders are already grappling with what AI-era credibility requires. The same precision that makes AI-powered compliance effective is the precision that makes a brand citable. Finance’s relationship with technology is already being redefined and the evidence layer is part of that redefinition.
Anthropic chief executive Dario Amodei has described powerful AI as potentially becoming a “country of geniuses in a datacenter”. That may sound distant from the practical work of communications. It is not.
If AI becomes a primary interface for information, then every company has to ask a basic question: what does the machine know about us?
Most cannot answer properly. They know their rankings. They know their organic traffic. They know their media mentions. But they do not know:
That gap is becoming dangerous.
The next phase of reputation management will not be about flooding the internet with content. It will be about creating a cleaner evidence layer:
It also means accepting a hard truth: brands do not control AI answers. No credible agency can guarantee a specific citation in ChatGPT, Claude, Gemini, Perplexity or Google AI Overviews. The systems change too often, the prompts vary too much, and the source selection process is not fully visible.
But brands can control the quality of the information available about them. They can make their claims clearer. They can publish better evidence. They can earn stronger third-party coverage. They can structure facts so they are easier to extract, quote and cite.
That is the real opportunity behind AI citations. Not gaming the system. Not buying mentions. Not pretending that generative engine optimisation is magic. Just making the public record around a company more useful.
For twenty years, marketers fought to be ranked. Now they must fight to be cited.
The companies that win will not be the ones with the most adjectives in their press releases. They will be the ones with the clearest evidence.
