
Rankings still rise while clicks fall, because Google now answers inside the results page. aeo aims at being the answer shown in an answer engine; GEO at being cited inside a generated answer. The two are not interchangeable, and Google's AI surfaces and the standalone assistants cite different sources, so the budget decision turns on where your buyers ask. As of November 2026 the two are usually sold together, and measured badly.
The pattern is structural, not a temporary dip. SparkToro's 2026 study, built on Similarweb's desktop and mobile panel for January to April 2026, found that 68.01% of Google searches in the United States ended without a click. In 2024 the same measure, from a different panel, stood at 60.45%. That is a rise of 7.5 percentage points in two years, which the authors call the fastest acceleration of the phenomenon in the last decade. Their research in 2019 had put the zero-click share at 49%.
The mechanism sits on the same page. AI Overviews, found on 20%+ of all searches, reduce click-through rate by nearly 60% when they appear. A page can hold position one and still lose the visit, because the summary above it answered the question already. Google's own adoption numbers explain why the feature keeps spreading: at its developer conference on 19 May 2026, Google reported AI Overviews at over 2.5 billion monthly active users and AI Mode past 1 billion monthly active users within a year of launch.
One caveat keeps the picture honest. Only 0.34% of searches made their way to AI Mode from January to April 2026, according to SparkToro. AI Mode is large in users and small in share of searches so far, and the shift toward it is still early. That matters when you allocate budget: the zero-click loss is real today, but the engine mix is mid-move.
The acronyms overlap in practice and diverge in what they promise. Working definitions help when you read a proposal, and Google's search team draws the line in much the same place as practitioners do: aeo is the search-feature game, GEO the generative-engine one.
| aeo | GEO | |
|---|---|---|
| Target | The answer block on an answer engine's results page | The citation inside an answer an engine generates |
| Typical surfaces | Google AI Overviews, AI Mode, featured snippets | ChatGPT, Claude, Gemini, Perplexity, and the AI layer of search |
| What is counted | Presence in the answer for a question | Mention, accurate description and source citation per answer |
| What you produce | Answer-shaped pages and structured facts | Citable knowledge content plus off-site presence |
| Failure mode | The answer appears, the page never does | The answer appears and describes you wrongly |
The distinction sounds academic until you measure it. A brand named in every Google AI Overview but never in ChatGPT has not won AI visibility; it has won half of it. An agency reporting one blended score will not show you which half, and a product change on either side can move the score without telling you where. Ask for the split.
Google's documentation supports the narrow reading. Search Central states that there are no additional requirements to appear in AI Overviews or AI Mode, and that the usual SEO best practices remain relevant. It also explains that both features may use query fan-out, issuing multiple related searches across subtopics and data sources, which is why one buyer question becomes several retrievals and several candidate sources.
A buyer who types a category question does not send one query to the engine. Under fan-out the engine runs related searches across subtopics and data sources, then builds the answer from what comes back. Your page competes with the whole cluster behind the question, not with one query.
For anyone buying aeo or GEO work, two consequences follow.
First, the question set you track must be the buyer's own questions, not your product names. A brand-name query tells you whether the engine knows you exist; a category question tells you whether it recommends you. Monitor both, and treat the same question run on a second engine as extra sample size rather than as a new question in the count.
Second, the fan-out list has to be read, not guessed. The follow-up queries an engine issues reveal the information need the first query only hints at. If the fan-out hunts for pricing, comparisons or local availability and your site keeps none of that in text, no optimisation on the head query will fix the answer.
This is where an aeo programme and a GEO programme stop being the same purchase. Search Engine Journal's 2026 study with AirOps, tracking roughly 380,000 AI answers across 81 enterprise software topics, found the citation split runs along the line between search features and standalone assistants.
| Domain | Google's AI surfaces | Standalone assistants |
|---|---|---|
| Vendor and product sites | 52% to 72% of citations, by surface | 72% on Gemini |
| 39% to 52% of citing answers | 21% on ChatGPT, absent on Claude | |
| YouTube | 38% to 47% of answers | 0.3% on Gemini; essentially none on ChatGPT and Claude |
| 9% to 13% of answers | 6% on Perplexity; none on ChatGPT, Gemini and Claude | |
| Review sites | 9% to 21% | About one in five citations on Claude and Perplexity; G2 in 30% of Claude's citing answers |
Read that table as two retrieval systems rather than one. AI Overviews and AI Mode ground their answers in live Google Search results and inherit what already ranks there, which is why the videos and Reddit threads that rank in Search show up in the answers. ChatGPT, Claude, Gemini and Perplexity run their own retrieval, lean on product pages and review profiles, and largely ignore social platforms.
The study's summary makes the split concrete: Google's two AI surfaces put 7% of their citations into social platforms, while ChatGPT, Gemini and Claude sit at 0.3% or below. A brand can be everywhere in Google's AI answers and invisible in ChatGPT, or the reverse. The diagnosis decides the purchase. If your buyers ask Google, build for the search-feature side; if they ask ChatGPT or Claude, build pages and review profiles.
The order below keeps each layer checkable before the next one starts. It follows the structure of the news study, restated as a sequence you can run, and applies to aeo and GEO work alike because the layers are the same.
Two checks decide what to build first: ask buyers on sales calls which AI tools they use, and pull your category's source mix before setting a budget. The study notes review citations range 3.3x across categories and community citations 2.5x, so a borrowed playbook can pour money into the wrong layer.
Take this to the next proposal meeting. A serious aeo or GEO proposal answers all nine points; a deck that skips most of them is selling a promise.
Two questions expose most weak proposals. "Which engine does this score come from?" separates a real measurement from a blended number. "What happens in the first monthly review if the mention rate does not move?" separates a scoped engagement from a retainer that renews on activity.
Topkee treats AI visibility as three separate measurements rather than one ranking: whether the brand is mentioned in valid answers, whether it is described accurately, and whether its own site appears as a source. Its GEO service runs a fixed buyer-question set, sampled by API across agreed AI engines, and delivers a monitoring report, a content pack with a ledger of every buyer question and target page, a priority action list and a monthly review. The active question count and the monthly fee scale with the tier, and the packages are documented in Hong Kong dollars for the Hong Kong market, so a Singapore client should ask for the local quotation.
The boundary matters as much as the scope. Topkee states plainly that it does not guarantee AI mentions, citations, fixed impressions, query volume, rankings or deals, and that API samples are not identical to what every consumer sees. On the Google side it uses Search Console's generative-AI performance report only as a supplement for AI Overviews and AI Mode impressions, and notes that whether the data is available depends on the client's account. GEO can be bought on its own or alongside SEO, with shared base content counted once.
For a reader choosing between aeo and GEO, that framing is the useful part. The question set, the engine list and the metric definitions are agreed before work starts, so the first monthly report shows where the brand stands on each "internet" instead of a single number that hides which one is missing. That is also how you tell a scoped engagement from a promise, whatever the proposal is called, and why a best geo agency singapore shortlist is worth reading against the nine-point checklist rather than against testimonials.
Before you move budget, read the AI visibility basics breakdown and compare it with your current reporting. If the report gives one ranking position per keyword, it cannot tell you whether AI answers cite you. Then look at how AI-driven discovery is reshaping demand in the region; the same surface-by-surface logic applies outside software categories, and the layer order seldom changes.
For the buying decision itself, Topkee's GEO service overview at aeo sets out the question counts, deliverables and metric definitions the checklist above asks for. It is the right page to open when you are comparing proposals for a best geo agency singapore shortlist.
No, though agencies often use the two interchangeably. aeo targets the answer block on an answer engine's results page, such as AI Overviews or a featured snippet. GEO targets being cited inside an answer generated by a model, which includes the standalone assistants. They share content work but are measured on different surfaces.
Not on this evidence. Google Search Central states that the usual SEO best practices remain relevant for AI Overviews and AI Mode, with no additional requirements. What changes is the expectation: ranking still shapes who gets cited, but it no longer guarantees a visit, and a second layer off-site is usually needed.
Trace it to your buyers. If they search Google, work on Reddit, YouTube and LinkedIn matters most, because Google's AI surfaces cite those domains far more than the standalone assistants do. If they open ChatGPT or Claude, your own pages and review profiles carry the weight. Ask on sales calls before you decide.
Look for a named question set, named engines, separate mention, description and citation metrics with denominators, and an explicit list of what is not guaranteed. A provider that gives one blended visibility score, or promises specific AI recommendations, has not shown you how it measures.
No. Google's own documentation states that indexing and serving are not guaranteed even when a page meets every requirement, and providers that follow the measurement discipline, including Topkee, state that AI mentions, citations, impressions and query volume are not guaranteed. What can be measured is the sample: the question set, the engines, the date and the share of valid answers in which the brand appears.