The next SEO battle is not rankings. It is AI citations

For thirty years, online visibility followed a familiar ritual. A buyer opened Google, typed a query, scanned a page of links, clicked through to a few websites, and made a judgement.

That ritual is not dead, but it is weakening. The search box is becoming a conversation. The list of links is becoming a generated answer. The buyer who once searched “best cybersecurity companies for ransomware protection” or “top negotiation training providers for enterprise teams” is increasingly asking ChatGPT, Claude, Perplexity, Gemini or Google AI Overviews to do the first round of research.

That changes the nature of visibility. The old question was whether a company ranked. The new question is whether an AI system can understand it, trust it and cite it.

The old question was whether a company ranked. The new question is whether an AI system can understand you, trust you and cite you.

This is the emerging world of AI citations, LLM citations and generative engine optimisation. The language is still settling, but the commercial direction is already clear. In an AI-mediated search journey, the brand that does not appear in the answer may never enter the buyer’s consideration set.

Google’s own numbers show that this is no longer a fringe behaviour. At Google I/O 2026, Sundar Pichai, the chief executive of Google and Alphabet, said AI Mode was “our biggest upgrade to Search ever”. Google says AI Overviews now has more than 2.5 billion monthly active users, while AI Mode has passed 1 billion monthly active users in its first year.

That is not a side product. It is Google rebuilding search around generated answers — and deciding, query by query, which sources are worth quoting back to the user.

OpenAI’s figures tell the same story from the other side of the market. A study by OpenAI’s Economic Research team and the Harvard economist David Deming analysed 1.5 million conversations and referred to ChatGPT’s 700 million weekly active users. It found that 49 per cent of ChatGPT messages fall into the category of “Asking”, where people use the system as an advisor rather than simply as a tool for writing or completing tasks.

For marketers, that distinction matters. If ChatGPT is being used as an advisor, then AI answers are becoming part of the decision-making layer. They influence which companies are discovered, which sources are trusted and which claims are repeated.

McKinsey has called AI search the “new front door to the internet”. Its analysis says half of consumers already use AI-powered search, that 44 per cent of AI-search users describe it as their primary and preferred source of insight, and that up to $750 billion in US consumer spend could flow through AI-powered search by 2028. McKinsey also estimates that 20 to 50 per cent of traditional search traffic could be at risk as AI captures decisions earlier in the journey.

That is the commercial fear now taking hold in marketing departments. A company may still have a website. It may still have a search strategy. It may even rank well on Google. But if buyers ask an AI assistant for a shortlist and the brand is missing, the click never happens.

The money moving into the sector confirms that AI visibility is becoming a serious category. Profound, a New York-based AI search platform, raised a $35 million Series B led by Sequoia in 2025, bringing its total funding to $58.5 million. The company says more than 2,000 marketers from over 500 organisations use its platform daily.

“AI assistants are becoming the new front door to every business. If you’re not showing up in those conversations, you’re invisible to billions of consumers.”

The point is not that every company needs another dashboard. It is that brand discovery is moving upstream. By the time a prospect lands on a website, the shortlist may already have been formed elsewhere.

That problem is particularly acute for companies that still treat SEO as a question of keywords and backlinks alone. The old search economy rewarded pages that could rank. The new AI search economy rewards information that can be used. Those are different disciplines.

A company homepage might say it is “a leading provider of innovative solutions”. A press release might announce a “strategic partnership to transform customer experience”. A case study might claim that a platform “delivered significant improvements”.

These phrases are familiar. They are also weak citation material. They do not contain a precise fact. They do not include a number. They are not easily attributable, and they are difficult to verify. Most importantly, they give an AI system very little to cite.

A stronger claim looks different. It says who did what, when, according to which source, with what measurable result. “Company X analysed 5,000 customer support tickets and found that 42 per cent related to onboarding confusion” is far more useful than “Company X helps businesses improve customer experience”. The first gives an AI system something concrete to work with. The second sounds like every other corporate claim on the internet.

The evidence suggests that AI visibility is its own discipline. A 2026 study of Google AI Overviews analysed 55,393 trending queries over a 40-day period. It found that AI Overviews appeared on 13.7 per cent of queries overall, rising to 64.7 per cent for question-form queries. Crucially, nearly 30 per cent of domains cited in AI Overviews did not appear in the standard first-page Google results.

That should worry any company that assumes normal SEO performance is enough. If AI Overviews cite sources that are not simply copied from page one of Google, then the AI citation layer has its own logic.

Another 2026 study of LLM brand reputation looked at 167,551 URL-grounded citations across 128 brands, 12 home markets and 13 languages. It found that 85.7 per cent of URL citations pointed to third-party sources, while only 14.3 per cent pointed to brand-owned domains.

85.7% of the URLs that AI systems cite point to third-party sources. What others say about you matters more than what you say about yourself.

This is where digital PR, journalism and search optimisation begin to overlap. In the old model, a press mention was useful because a person might read it, remember the brand and click through. In the new model, that mention can also become machine-readable evidence. It can help an AI system understand that a company exists, what it does, who has written about it and whether its claims are supported by public sources.

That does not mean every article becomes an AI citation. Thin, promotional coverage is still weak, and vague press releases are still vague. But a clear article with named people, specific claims, dated evidence and attributed statistics is far more valuable than a generic brand page full of adjectives.

If AI systems increasingly depend on source-backed information, then credible reporting becomes part of the infrastructure of discovery. New Statesman Canada has already covered one version of this shift in retail, where AI agents are beginning to compare products before the consumer reaches the shopfront. The same logic now applies to almost every knowledge market. Search is not just a traffic source. It is becoming a judgement system.

Canada is not insulated from this. Mark Carney’s “AI for All” strategy has already framed artificial intelligence as a national productivity issue, with more than $2bn of federal money, a promised 250,000 jobs by 2031 and a target of lifting business AI adoption from 12 per cent to 60 per cent by 2034.

But adoption is only one side of the problem. Canadian firms will not simply use AI — they will also be judged by it. Banks, law firms, software companies, universities, retailers, consultancies and public bodies will all be described, compared and summarised by systems they do not control. If those systems rely on old data, thin content or poorly attributed claims, the resulting answer may be incomplete, outdated or wrong.

Anthropic’s Dario Amodei has argued that AI policy is moving too slowly for the pace of the technology. In a 2026 policy note, Anthropic wrote that “AI is advancing at exponential speed, and the policymaking process was built for a slower world”. Amodei has also used the phrase “a country of geniuses in a datacenter” in his essay to describe the possible scale of powerful AI systems.

That may sound abstract, but for businesses the practical question is immediate: what does the machine know about you? Where did it learn it? Which sources does it trust? Which competitors does it mention instead?

Most companies cannot answer those questions with confidence. They can say where they rank for a keyword, how much organic traffic they received last month and which pages converted. But they often cannot say how ChatGPT describes them, whether Perplexity cites them, whether Claude understands their category, or whether Google AI Overviews includes their evidence when summarising the market.

Some of the market will inevitably become noisy. There will be agencies promising to “rank in ChatGPT”, dashboards presenting confidence where there is still uncertainty, and vendors implying that AI systems can be gamed the way search engines once were. Brands should be sceptical of all of it. No serious company can guarantee that a specific AI system will cite a specific brand in a specific answer. Models change, retrieval systems change, prompts vary, and source selection is unstable.

But that does not mean companies are powerless. They can make their information easier to understand. They can publish clearer facts, replace empty claims with attributed evidence, build better founder and expert profiles, and make sure product and company information is consistent across trusted third-party sources. They can track how AI systems describe them and correct the obvious gaps.

This is where citation infrastructure comes in. citations.press describes itself as a structured citation index for the AI search era, publishing source-backed facts where each entry names who said it, links to the original source and is reviewed before publication, in both human-readable and machine-readable formats.

The wider point is not about any one platform. If AI systems are becoming the interface through which people interpret information, then brands need to think less like advertisers and more like sources. That means answering basic questions properly. Who are you? What do you do? What evidence supports the claim? Who said it, and when? Is there a number, a named person, a reliable source — and can the fact be extracted without wading through 800 words of marketing language? If the answer is no, the brand is not ready for AI search.

For marketers, the temptation will be to turn this into another acronym. SEO became AEO, AIO, GEO and LLM visibility. The labels matter less than the underlying shift: the internet is moving from pages to answers, and in that environment visibility is not simply about being found. It is about being selected as evidence.

The companies that adapt early will not be the loudest. They will be the clearest — fewer empty claims, more source-backed facts, data that is specific, current and attributable. They will treat external validation as seriously as they once treated rankings, and they will watch how AI systems describe them with the same attention they once gave to a page of Google results.

The companies that wait will find out the hard way. They will not lose ground because an AI system dislikes them. They will lose it because, when a buyer asks, the system cannot explain them clearly — and so it reaches for a competitor it can. In an internet that increasingly answers rather than lists, the prize is no longer simply being found. It is being trusted enough to be cited.