When search becomes conversation conversational commerce starts before your website

I

Customers no longer search, they ask

In a study published in July 2025, the Pew Research Center analyzed 68,879 real Google searches carried out in March 2025 by 900 American adults¹. Queries of one or two words triggered an AI-generated summary 8% of the time. Those of ten words or more triggered one 53% of the time, and those phrased as a question (who, what, when, why) 60% of the time¹.

These figures describe a simple mechanism. The more a request looks like a sentence, the more the engine answers itself instead of offering a list of links. Google confirms this on its side: announcing the arrival of AI Overviews and AI Mode in France on July 22, 2026, the company says that on average, “a search in AI Mode is three times longer than a traditional search,” referring to data published for the US market².

The same shift appears on the conversational assistant side. A study published in September 2025 by researchers from OpenAI, Harvard and Duke, based on a sample of about 1.1 million conversations, classifies about 49% of the messages sent to ChatGPT in the “Asking” category: the user is looking for information or advice to make a decision³. The share of conversations devoted to seeking information rose from 14% to 24% of usage between July 2024 and July 2025, and the authors describe this use as a very close substitute for web search³.

The keyword was a concession made to the machine. A need (“I’m looking for a lightweight waterproof jacket for a trek in Scotland in April”) was translated into a string of terms the index could understand (“lightweight waterproof jacket”). Answer engines remove this constraint. Customers now state their need the way they think it, with their context, their constraints and their hesitations.

II

A generation that phrases its needs in sentences

In France, the 2026 edition of the Digital Barometer, carried out by CREDOC in June 2025 among 4,145 people aged 12 and over, finds that 48% of the population uses generative AI, 15 points more in one year⁴. Use reaches 85% among 18-24 year-olds and 76% among managers and professionals. Searching for information is the most frequently cited use, by 73% of users⁴. According to the same press release, generative AI reached this level of adoption in three years, when the smartphone had taken seven⁴.

In the United States, a Pew Research Center survey conducted in February 2026 among 5,119 adults indicates that 49% of them use AI chatbots (the survey’s term), and that 42% use them to look up information⁵. The OpenAI study, for its part, notes that nearly half of the messages sent by adults to ChatGPT come from users under 26³.

What follows is analysis. A growing share of your customers, and in particular the youngest, are getting used to receiving an answer tailored to their situation rather than a list of options to sort through. This habit doesn’t stay confined to the search engine. It becomes a general expectation: to be understood in their own words, to be able to clarify, go back, compare by asking a second question. A visitor who arrives on your website after talking with an answer engine doesn’t spontaneously fall back on the reflex of menus, filters and category pages. They look for the place to ask their question.

III

Answer engines have changed scale

The idea of conversational search is no longer experimental. It now reaches audiences of several hundred million people.

OpenAI launched ChatGPT search on October 31, 2024, with answers accompanied by links to sources, then opened it on February 5, 2025 to all users in the regions where ChatGPT is available⁶. On February 27, 2026, the company announced more than 900 million weekly active users for ChatGPT⁷.

Google built the generated answer directly into its results page. At its I/O conference on May 19, 2026, Sundar Pichai said that AI Overviews had more than 2.5 billion monthly active users, that AI Mode had passed one billion monthly active users in a year, and that the Gemini app exceeded 900 million monthly active users⁸. In France, these features have been rolled out since July 22, 2026². AI Mode there relies on a technology Google calls “query fan-out”: the question is broken down into sub-topics and several searches are launched simultaneously².

Specialized answer engines such as Perplexity add to these players. What these services have in common is structural: the answer comes before the click. The website is no longer the default destination of the search; it becomes one source among others, consulted by the engine before, possibly, being consulted by the user.

IV

Zero click: what the studies measure

The effect on website traffic is measured by several independent studies. They converge on the trend, with different methods and orders of magnitude.

In the July 2025 Pew study, users clicked on a traditional link in 8% of visits when an AI summary was displayed, versus 15% when there was none¹. They clicked on a link cited in the summary itself in only 1% of cases. They ended their browsing session after the results page in 26% of cases with a summary, versus 16% without¹.

Ahrefs published an update of its analysis of 300,000 keywords in February 2026. The presence of an AI Overview is correlated there with an average click-through rate 58% lower for the page ranked first, in December 2025⁹. This figure compares the observed click-through rate with a projection built from informational keywords without an AI Overview: it is an estimate, not a controlled experiment⁹.

Seer Interactive tracked 3,119 informational queries across 42 client organizations. On queries showing an AI Overview, the organic click-through rate fell from 1.76% in June 2024 to 0.61% in September 2025¹⁰. The study also observes that brands cited in the AI Overview get more clicks than others, while stating that it cannot prove the citation is the cause¹⁰.

Google disputes the most pessimistic reading of this data. In a post published on August 6, 2025, Liz Reid, head of Google Search, states that the total volume of organic clicks sent to websites has remained “relatively stable” year over year, and that the average quality of clicks has increased¹¹.

These positions aren’t incompatible. Third-party studies mostly measure the click-through rate, often on informational queries, where the generated answer frequently suffices. Google speaks of overall volume, which includes the growth in the number of searches. For a decision-maker, the prudent conclusion is this: on simple questions, a growing share of answers is given without a visit; the clicks that remain are, according to Google, of better quality, which is consistent with an intent already refined upstream.

E-commerce data points the same way. According to Adobe Analytics, traffic to US retail sites from generative AI tools grew 693.4% year over year during the 2025 holiday season, from November 1 to December 31, based on more than 1 trillion visits analyzed¹². Adobe notes that the user base remains modest¹². The growth is therefore strong, from a still low level.

V

From the answer engine to conversational commerce

The term “conversational commerce” was popularized by Chris Messina, notably in a January 2016 article where he used it to describe the use of chat, messaging or voice to interact with brands and services¹³. Ten years later, the subject is no longer limited to messaging. Conversation becomes the very interface for discovery, advice and sometimes purchase.

The big platforms are building the pieces of this journey. On January 11, 2026, Google introduced the Universal Commerce Protocol, an open standard for agentic commerce co-developed with Shopify, Etsy, Wayfair, Target and Walmart, with payment initially limited to eligible US merchants¹⁴. On October 15, 2026, Google launched Universal Cart in the United States, a cart available in Google Search and the Gemini app, which lets people save products from several merchants and then pay with Google Pay at many merchants or complete the purchase on their site¹⁵. Both announcements concern the US market.

What follows is a reasoned hypothesis. The buying journey is now split into two stages. The first takes place on surfaces you don’t control: the customer refines their need, compares, eliminates. The second begins when they arrive on your site. By then, they have already stated their request, often several times. If they find a site organized like a catalog, with a search bar that expects exact keywords, they have to backtrack and relearn a mode of interaction they have left behind.

Your website must therefore, in turn, become a place where people can ask. In practice, this means three things. Natural-language search, able to understand “a gift for an 8-year-old who loves astronomy, under €50.” An answer available where the question arises, in particular on the product page (“is this compatible with my model?”, “which size should I choose?”). And a conversation able to act: track an order, book an appointment, open a request, or hand off to a human advisor with the full context. Conversational commerce isn’t an additional channel. It’s the continuity between how the customer searched and how they are welcomed.

VI

Being visible when the answer comes before the click

The first finding is reassuring. In its documentation for site owners, updated in December 2025, Google states that there are “no additional requirements” and no special optimization to appear in AI Overviews or AI Mode¹⁶. A page must be indexed and eligible to be shown with a snippet. Google also specifies that no specific schema.org markup or dedicated AI text file is needed, and that structured data must match the visible content of the page¹⁶.

The vocabulary has changed, though. People talk about GEO (Generative Engine Optimization) or AEO (Answer Engine Optimization) to describe the work of visibility in generated answers. The term GEO was formalized in a research paper presented at the KDD 2024 conference, which measured up to a 40% visibility gain on an experimental benchmark, with effectiveness varying by domain¹⁷. This result applies to a test environment. It is not a guarantee applicable to commercial engines, whose source-selection mechanisms are not public.

What can be stated with caution comes down to a few principles, consistent with Google’s recommendations¹⁶:

• Answer real questions. Queries are getting longer and taking the form of questions¹. Content that explicitly addresses a question, with a clear answer in the first lines followed by development, matches this demand.
• Cover the sub-questions. Since the engine breaks a request down into several searches², content that covers the topic from its different angles (use, compatibility, care, comparison) has more chances of being consulted.
• Keep the essentials as text. Google recommends keeping important information in text form and keeping Merchant Center and Business Profile listings up to date¹⁶.
• Align structured data with visible content. Markup helps engines understand a page, provided it describes what the user actually sees¹⁶.
• Decide on your exposure. Google has opened a control in France within Search Console that lets site owners choose whether their site appears in these generative AI features². It’s a trade-off to make with full knowledge, not a default setting.

None of these practices guarantees a citation. They increase the likelihood that your content will be understood, retained and judged relevant. They also have an indirect effect: a website that answers its customers’ questions clearly is better prepared to welcome them into the conversation.

VII

What this means for your organization

The shift to conversational search affects acquisition, the website and customer relationships all at once. It calls for a four-step approach.

Measure first. In your analytics tools, isolate the traffic coming from AI answer engines and track its conversion rate separately. Compare how your clicks and impressions evolve in Search Console on your informational queries. You’ll then know whether zero click affects you, and on which pages.

List your customers’ questions. The requests sent to your customer support, the exchanges with your advisors and the internal searches on your website contain the questions as your customers phrase them. This corpus is the raw material for your content and your future conversation.

Make the website conversational where questions arise. Start with the points where the visitor hesitates: internal search, high-value product pages, contact. Replacing a keyword search bar with natural-language search, or adding an answer on the product page, lets you measure the effect without redesigning the whole site.

Plan the human handoff. A conversation that doesn’t know how to answer must pass the request, with its context, to your teams. That’s what lets you open the conversation to new topics knowing your teams stay in control.

VIII

Where QLOA stands

IX

Frequently asked questions

What is conversational commerce?

Conversational commerce refers to all the commercial exchanges carried out in a conversation, written or spoken, rather than through menus and forms. The customer expresses their need in their own words, receives tailored advice, can compare, ask a follow-up question and, depending on the case, book, order or be connected to an advisor. The term was popularized by Chris Messina in the mid-2010s and is taking on a new dimension today with AI agents.

What is a zero-click search?

A zero-click search ends without the user clicking on a link to a website. The answer is displayed directly on the results page, for example in an AI Overview. According to the Pew Research Center (2025), users click on a traditional link in 8% of visits when an AI summary appears, versus 15% without a summary. The size of the effect varies by type of query.

Do AI Overviews reduce website traffic?

Several third-party studies measure a drop in click-through rate on informational queries showing an AI Overview: Ahrefs estimates it at 58% for the first position in December 2025, by correlation. Google, for its part, states that the total volume of organic clicks has remained relatively stable year over year (August 2025 post). The effect therefore depends heavily on the nature of your pages and your queries.

How do you appear in ChatGPT or Google AI Mode answers?

Google states that no special optimization is required for AI Overviews and AI Mode: the page must be indexed and eligible for a snippet. Best practices remain those of search optimization: content that clearly answers real questions, important information in text, structured data consistent with the visible content. No method guarantees being cited.

What is the difference between SEO, GEO and AEO?

SEO aims at ranking in traditional search results. GEO (Generative Engine Optimization) and AEO (Answer Engine Optimization) aim at being present in the answers written by AI answer engines. All three rely largely on the same fundamentals. The difference lies mainly in the goal: being cited as a source for an answer, and not just being ranked among links.

Why make your website conversational if traffic is declining?

Because the visitors who still arrive have often already refined their need in an answer engine. They’re used to asking rather than navigating. Natural-language search or an answer available on the product page extends this habit instead of interrupting it. Every visit matters more when their number shrinks on simple queries.

X

Sources

1. 1

   Pew Research Center, “Google users are less likely to click on links when an AI summary appears in the results,” July 22, 2025. pewresearch.org

2. 2

   Google France (Sébastien Missoffe), “Launch of AI Overviews and AI Mode in Google Search in France,” July 22, 2026. blog.google

3. 3

   A. Chatterji, T. Cunningham, D. Deming et al. (OpenAI, Duke University, Harvard University), “How People Use ChatGPT,” September 15, 2025. cdn.openai.com

4. 4

   Arcep, Arcom, CGE, ANCT (study carried out by CREDOC), “Digital Barometer: half the population uses generative AI,” press release of February 9, 2026 (Digital Barometer, 2026 edition). arcep.fr

5. 5

   Pew Research Center, “Americans and AI 2026: Chatbots, smart devices and views on impact,” June 17, 2026. pewresearch.org

6. 6

   OpenAI, “Introducing ChatGPT search,” October 31, 2024 (updated February 5, 2025). openai.com

7. 7

   OpenAI, “Scaling AI for everyone,” February 27, 2026. openai.com

8. 8

   Google (Sundar Pichai), “I/O 2026: Welcome to the agentic Gemini era,” May 19, 2026. blog.google

9. 9

   Ahrefs (Ryan Law, Xibeijia Guan), “Update: AI Overviews Reduce Clicks by 58%,” February 4, 2026. ahrefs.com

10. 10

    Seer Interactive (Tracy McDonald), “AIO Impact on Google CTR: September 2025 Update,” November 4, 2025. seerinteractive.com

11. 11

    Google (Liz Reid), “AI in Search is driving more queries and higher quality clicks,” August 6, 2025. blog.google

12. 12

    Adobe, “Adobe: Holiday Shopping Season Drove a Record $257.8 Billion Online with Consumers Embracing Generative AI Tools,” January 7, 2026. news.adobe.com

13. 13

    Chris Messina, “2016 will be the year of conversational commerce,” Medium, January 19, 2016. medium.com

14. 14

    Google, “New tech and tools for retailers to succeed in an agentic shopping era”, January 11, 2026. blog.google

15. 15

    Google (Suresh Ganapathy), “An intelligent new shopping cart”, October 15, 2026. blog.google

16. 16

    Google Search Central, “AI features and your website,” updated December 10, 2025. developers.google.com

17. 17

    P. Aggarwal, V. Murahari, T. Rajpurohit, A. Kalyan, K. Narasimhan, A. Deshpande, “GEO: Generative Engine Optimization,” KDD 2024. arxiv.org