Every week, a new headline pops up announcing the death of SEO. In 2026, the villain has a name: Google's AI Overviews and AI Mode. And the numbers are frightening. A study by Seer Interactive, covering 5.47 million queries, measured a 61% drop in CTR when an AI Overview appeared. The aggregated organic traffic of major publishers is 42% below its pre-AI level. HubSpot reportedly lost between 70% and 80% of its organic traffic. Business Insider, close to 55%.
It's tempting to read this data and conclude that Google is stealing your traffic. It's the comfortable narrative. It's also the wrong one.
The same studies that document the collapse hide a second piece of data that almost no one mentions: brands that are cited within the AI Overview receive 35% more clicks, and brand searches with an AIO present increase their CTR by 18%. We are not facing the death of traffic. We are facing its redistribution. And the question is no longer "how do I get to position 1?", but "why does AI cite my competitor and not me?".
From the ranking economy to the citation economy
For twenty years, SEO functioned as a position auction. Ten blue links, one winner per query, and a known CTR curve: the first result took 30-40% of clicks, the second half of that, and from the fifth almost nothing. The entire industry optimised for the same goal: to move up that list.
Generative engines break that logic. ChatGPT, Perplexity, Claude, and AI Overviews don't return a list of ten options for the user to choose from. They return a synthesised answer and cite, on average, between 2 and 7 sources. Search no longer distributes attention among ten pages: it concentrates it on the handful of domains that the model decides to use as a source.
This changes the game entirely. In the ranking economy, coming seventh still gave you crumbs of traffic. In the citation economy, coming seventh means not existing: if the model chooses three sources and you are not one of them, your visibility for that query is zero, no matter how good your content is "on paper". There's no consolation prize. Either you get cited, or you disappear from the conversation.
SEO is not dead. It has levelled up. The goal is no longer to rank a URL, it's to be the source that AI deems worthy of citation.
That's why the 42% drop and the 35% rise are two sides of the same coin. The traffic lost by those who optimised for the blue link is gained by those designed to be cited. It's a near zero-sum game, and it's being redistributed right now. Those who understand the new rules in 2026 will take the share that others are giving up due to panic or inertia.
Why a citation is worth more than the click you lost
When an executive sees the 42% drop, they focus on the clicks that are no longer coming in. But there's a second effect, slower and more valuable, that doesn't appear in the traffic report: trust transfer. When ChatGPT answers a user and says "according to ZDS...", it's lending its credibility to that brand. The user hasn't clicked, but they've registered the name in a context of authority. They've received an implicit recommendation from a source they trust.
That registration pays off later. This is exactly what explains the fact that brand searches with an AIO present increase CTR by 18%: someone saw your name cited by AI, remembered it, and searched for you directly again. An informational click you lost in the generic answer becomes, weeks later, a brand search with much higher intent. The funnel didn't break: it lengthened and became harder to measure with traditional tools. Anyone who only looks at direct traffic the next day will never see that chain. Anyone who measures citation share will.
There's a common objection: "Okay, but that AI traffic doesn't convert." Sometimes it's true, and for a specific reason: most citations occur in informational queries at the top of the funnel. But therein lies the opportunity. Being the source that educates the user during the discovery phase is what positions you as a reference when that same user, months later, reaches the decision phase. The brand that taught wins the sale. The one that only appeared in a blue link that no one clicked, does not.
Four engines, four different criteria
The most expensive mistake we see in 2026 is treating "AI" as one thing. It's not. Each engine has a different citation temperament, and the data confirms it: only 11% of domains cited by ChatGPT also appear in Perplexity. Winning in one doesn't guarantee anything in the other. Optimising "for AI" in the abstract is like optimising "for the search engine" without knowing if you're talking about Google or Baidu.
ChatGPT: encyclopaedic and Wikipedia-like
ChatGPT rewards content that looks like an encyclopedia entry: clear structure, neutral tone, verifiable statements, and a strong signal of authority (E-E-A-T with a real author and bio). Its retrieval index relies heavily on Bing, so a step as unglamorous as submitting your sitemap to Bing Webmaster Tools has a direct impact. Content in question-and-answer format, short and extractable paragraphs, and concrete data with figures: that's what ChatGPT cites.
Perplexity: recency and community
Perplexity is the engine of current events. It prioritises recent sources and has a well-documented weakness for Reddit and communities. If your content strategy is to publish a "definitive" guide once and forget about it for two years, you are invisible to Perplexity. Here, frequency, visible update dates, and organic presence in conversations where your category is truly discussed win.
Claude: authority and intellectual honesty
Claude tends to cite sources with real authority and penalises inflated marketing. Content that acknowledges nuances, admits trade-offs, that doesn't sell smoke: that's what fits its criteria. Intellectual honesty, which was irrelevant in old SEO, here becomes a citation signal.
Google AI Overviews: uneven coverage by sector
The AIO does not appear equally everywhere. It covers 88% of health queries and 83% of education queries, but only 37% of entertainment queries. If you operate in a high-coverage vertical, the citation game is no longer optional: it's where the match is decided.
What really works to get you cited
The good news is that Generative Engine Optimisation (GEO) is not black magic. It's a set of concrete practices, many of which are the natural evolution of good SEO. This is what we move first when working on a client's AI visibility:
- Content designed for extraction. Models don't lovingly read your page: they scan it looking for quotable fragments. Concrete statements with numbers ("reduces CPC by 23%", not "greatly reduces cost"), clean semantic structure, question-and-answer patterns, and self-contained paragraphs that make sense out of context.
- Tier 1 Placements. Engines disproportionately trust a handful of publications: Forbes, TechCrunch, Reuters, WSJ, and leading industry publications. A mention in one of those sources carries more weight in your AI visibility than fifty directory links. Digital PR ceases to be a brand tactic and becomes citation infrastructure.
- E-E-A-T with a name and surname. Real author, verifiable bio, credentials, recognisable entity. AI cites people and organisations with identity, not "the editorial team".
- Proprietary data. The most reliable way to get cited is to be the original source of data. Studies, benchmarks, surveys, figures that no one else has. If your content is a rehash of what others have already said, the model will cite the original, not you.
- Structured data and visible freshness. Schema.org (FAQPage, HowTo, Article with author and date), real update dates, and a publishing rhythm that tells Perplexity you're still active.
And what you should stop doing
Optimising for the citation economy also means letting go of habits that no longer pay off. Continuing to invest the entire budget in fighting for the organic position 1 for purely informational queries, in verticals where the AIO answers 80% of the time, is throwing money away: even if you come first, the user already has their answer before scrolling down to the links. Producing generic content without proprietary data is also useless, because it doesn't provide anything citable. And measuring success only with classic Google ranking is like driving by looking in a rearview mirror that no longer reflects the road.
The new KPIs: the scoreboard changed language
If the goal changes, the metrics must change with it. Continuing to measure success with average Google position and the number of keywords in the top 3 is measuring yesterday's game. These are the indicators that truly matter in the citation economy:
- Citation share (AI share of voice). Out of every 100 relevant responses in your category, how many times do you appear compared to your competitors? It's the modern equivalent of market share in the search engine.
- Appearance rate per engine. Because winning in ChatGPT is not winning in Perplexity. You need the breakdown, not an average that hides the fact that you are invisible in half the market.
- Position within the answer. Being the first cited source is not the same as being the fifth. The first citation gets the attention and the link; the last ones, almost nothing.
- Sentiment. Being cited is good; being cited as a positive reference is much better. A model can mention you as an example of what not to do, and that also needs to be detected.
A real example of what this looks like: an industrial brand came to us convinced that "they were doing well on Google". Their average position was solid. When we measured their AI visibility, they appeared in 9% of relevant responses in their category; their main competitor, in 41%. In the classic search engine, they were on par. In the conversation that truly influences B2B purchasing decisions, their competitor was burying them. That gap is not visible in any traditional SEO tool. And you cannot close what you cannot see.
You cannot optimise what you do not measure
Here's the uncomfortable problem. Most companies have no idea if AI cites them, how often, in what position within the answer, or against which competitors. The Google Search Console panel doesn't say. Analytics doesn't say. You are competing on a field whose scoreboard you cannot see.
That's why at ZDS we built our own AI Visibility Tracker tool: we monitor how many times a brand appears against its competitors in ChatGPT, Perplexity, and Google AI Overviews responses, with what sentiment and in what position. While other agencies use generic third-party tools, we built our own because the citation economy is too important to manage blindly. It's not theory: it's the scoreboard of a game that is already being played.
The conclusion no one wants to headline
AI Overviews are not killing traffic. They are rewarding those who adapt and punishing those who insist on playing by 2015 rules. The 42% that is lost does not disappear: it changes hands. In 2026, your brand's visibility does not depend on where your link appears, but on whether artificial intelligence considers you a worthy source to cite. That is the new auction. And there is still room at the top, because most of your competitors are still optimising for a search engine that no longer exists.
The question is not whether SEO is dead. It's whether you will be a source or a footnote.
