The blue link isn’t dead, but it’s on a diet<\/h3>\n
First, the scale of this. Google’s AI Overviews have reached roughly 2.5 billion users, with AI Mode passing 1 billion monthly users<\/a>. That’s not a niche feature you can ignore until next year. It’s the default experience for a huge chunk of the people searching for what you sell.<\/p>\n And it changes the maths of clicks. Recent analysis pegs the click loss from AI Overviews at around 39.8% \u2014 with the clicks you lose being no lower-quality than the ones you keep<\/a>. Read that twice. It’s not just tyre-kickers vanishing; it’s genuine prospects getting their answer without ever landing on your page.<\/p>\n So the old scoreboard – position one, position two, position three – is quietly being replaced by a new one: citation rate<\/strong>. How often does the AI pick you<\/em> as a source worth pointing to?<\/p>\n Here’s the part that trips people up. Getting cited is not the same as getting the business.<\/p>\n Lily Ray’s 2026 B2B software study found something delightfully awkward: AI Overviews would happily cite self-promotional “best of” listicles, yet exclude the publisher itself from the actual recommendation 69% of the time<\/a>. In other words, your page can be used<\/em> to build the answer while a competitor gets named<\/em> as the answer. As Ray puts it, being cited is not the same as being chosen \u2014 and that distinction is the whole ballgame for credible AI-search measurement.<\/p>\n Peter Fox, our Head of Dev Team, has been chewing on exactly this:<\/p>\n “Stop treating an AI citation as the finish line. A citation means the model trusted your page enough to read it<\/em>. Whether it then recommends you comes down to structured, unambiguous signals – clean entities, schema, consistent facts across the web. We treat every important page as if a machine has to defend it out loud.”<\/p>\n<\/blockquote>\n That reframing matters, because it tells you where to spend your effort: not on gaming a mention, but on being the clearest, most trustworthy source in the room.<\/p>\n The instinct is to optimise your own site harder. Useful, but only half the story. Aleyda Solis’s July 2026 SaaS analysis found that most AI citation weight comes from outside<\/em> a brand’s own website<\/a> – third-party reviews, mentions, comparison pages, forums, the places your customers already hang out.<\/p>\n That should feel familiar to anyone who’s done real reputation work. AI models are essentially reading the room. If the wider web keeps associating your brand with a topic, the model treats that as a strong signal. If the only place singing your praises is your own homepage, it reads exactly as convincingly as it sounds.<\/p>\n Enough theory. Here’s the doable list.<\/p>\n 1. Audit your citation rate, not just your rankings.<\/strong> Pick your 20 most valuable queries and run them through AI Overviews<\/a> and AI Mode manually (or with a tracker like Semrush’s or Ahrefs’ AI features). Note two columns: were you cited?<\/em> and were you recommended?<\/em> The gap between those columns is your to-do list.<\/p>\n 2. Answer the question in the first two sentences.<\/strong> AI models love a page that states the answer plainly up top, then supports it. Bury your conclusion under 600 words of throat-clearing and you’ve made yourself hard to quote. Lead with the payoff.<\/p>\n 3. Feed the machine structured facts.<\/strong> Clear headings, FAQ blocks, definition-style sentences, and proper schema markup<\/a> all make your content easier to parse and cite. If a fact is important, state it as a standalone sentence a model can lift cleanly.<\/p>\n 4. Build presence off<\/em> your site.<\/strong> Earn mentions on the comparison pages, review sites, and industry publications your buyers already read. This is where a lot of citation weight is decided \u2014 and, conveniently, it overlaps almost perfectly with good old-fashioned link building and digital PR.<\/p>\n 5. Keep your facts consistent everywhere.<\/strong> If your pricing, product names, or founding date differ across your site, your LinkedIn, and Wikipedia, you’ve handed the model a reason to doubt you. Consistency reads as trustworthiness.<\/p>\n 6. Watch Search Console’s new AI controls.<\/strong> Google added tools for site owners to influence how content shows up inside AI features<\/a> \u2014 worth a look before you assume you have no levers.<\/p>\n![\"a](\"https:\/\/3way.social\/blog\/wp-content\/uploads\/2026\/07\/2026-geo-playbook.png\")
<\/a><\/p>\nCited vs. chosen: the distinction that ruins everyone’s day<\/h3>\n
\n
Where AI actually gets its confidence from<\/h3>\n
Practical things you can actually do this week<\/h3>\n
The mindset shift<\/h3>\n