Ranked, but not chosen.

For most of the last two decades, a strong Google ranking was one of the more dependable assets in marketing. It took effort to build, it held for a while, and it converted into traffic with reasonable predictability. Many organisations still hold those rankings. What they are losing is the visit that used to follow.

A 2025 study tracked close to 69,000 real Google searches by US adults. When an AI summary appeared at the top of the page, users clicked on a traditional result in 8% of visits. Without a summary, the figure was 15%. Links inside the summary itself were clicked in about 1% of visits. Longer, natural-language searches, especially those phrased as questions, were the most likely to trigger a summary.

That last finding explains why rankings and visibility are drifting apart. Keyword strategy was built on short queries: a category term, a product name, a modifier. People now type, or speak, the full question. AI systems answer it directly, drawing on whatever content addresses the situation most clearly. A page optimised for the keyword can keep its position and still be absent from the answer.

I see this in a current project where the audience is citizens rather than customers. They rarely search for an organisation's name plus a keyword. They ask what will happen to school renovations in their city, or how a transport plan affects their commute. Whether a position appears in the AI answer depends on whether any page answers that exact question. The same logic applies to every company whose customers arrive with a problem instead of a product name.

Christensen and colleagues argued in Harvard Business Review in 2016 that customers hire products to get a job done in a specific circumstance, and that the circumstance matters more than customer attributes. A question typed into an AI assistant is usually a job stated in full: the situation, the constraint, the outcome needed. Content that maps to jobs is the content these systems can use.

Aggarwal and colleagues showed at KDD 2024 that the way source content is written changes how often generative engines cite it. Adding sources, quotations and statistics raised visibility in their experiments by up to 40%. Generative systems favour content they can extract and justify.

When an organisation is missing from the answer, the cause usually falls into one of three categories:

• The page was written around a keyword and never states an answer
• The answer exists but is buried in marketing copy that a system cannot cleanly extract
• Or there is no page for the job at all, because the content plan was built from a keyword list rather than from what people ask

Collect the ten to twenty questions your sales and service teams hear most often, in your audience's own words. Run them through ChatGPT, Perplexity and Google's AI Mode. Record whether your brand appears, whether it is recommended, and whether your own pages are cited as sources. Then compare that list with the keywords your SEO plan is built on. The overlap is usually smaller than teams expect.

The fix is editorial first. One question per section, a direct answer in the opening sentences, evidence placed next to every claim, and structured data so the page can be parsed. Technical SEO remains necessary, but it is no longer sufficient.

Rankings, impressions and clicks describe the part of the funnel that still runs through the search results page. They say nothing about the part that now ends in an AI answer. Citation share, meaning how often your brand and your pages appear in AI answers to the questions that matter, belongs on the same dashboard. Until it is there, a growing share of the buying journey stays invisible to the people responsible for growth.