Who Recommends You When AI Answers? Publishers and Voucher Codes

Abstract

For years the internet ran on a simple loop: Google sends visitors to websites, the websites earn money from them, and that money pays for new content. Answers written by artificial intelligence (AI) break the loop, because people get what they need without opening a page. Several studies point to falling click rates [1][2][3][4]. At the same time, AI systems still need current, accurate information, and they find it with publishers: deal sites, newsletters, creators and guides [6][7]. For direct-to-consumer (D2C) brands that sell straight to shoppers, this changes the central question. It is no longer "How do I rank on Google?" but "Who recommends me when people and AI look for the best offer?" This article sets out what has changed, separates documented facts from our own judgement, and explains why exclusive, permanent voucher codes held with publishers are an effective lever for D2C shops in the UK.

1. Introduction: the situation and the problem

You will recognise the pattern from your own shop. Free visits from Google search are less dependable, and paid ads keep getting dearer. A third force has arrived. More and more people put their question straight to an AI assistant such as ChatGPT, Gemini or Perplexity, or they read the AI summary at the top of Google's results. These assistants run on large language models (LLMs): programs trained on vast amounts of text so that they can answer questions in full sentences.

For a shop this is more than a technical footnote. If the answer is already on the screen, nobody clicks through ten results. If you are not in the answer, you do not get the sale. This article asks three questions:

1. What is falling away from the old model?
2. Where do AI systems get their information, and what part do publishers play?
3. What does that mean for how shops win visibility and sales?

We draw on published studies and industry reports. Where the evidence is thin, we say so and label the statement as our assessment.

2. Background: how the old internet model worked

The classic model has three steps:

• A search engine delivers visitors. Someone asks a question, Google lists ten results, and they click one.
• Websites earn money. Sites fund themselves through advertising, subscriptions or commission on the sales they generate.
• New content appears. Newsrooms, creators and guide publishers use that income to pay for their work and publish more, which in turn produces more search results.

Shops benefited too. Good rankings brought free traffic, and guides and comparison sites passed on people who were ready to buy. Affiliate marketing, where a partner earns a commission when a purchase comes through their recommendation, is built on this loop. We have no comparable, independently verified market figure for the UK that we can cite here. For Germany, the industry body BVDW reports that affiliate links brought in €18.7 billion of sales in 2025, around 16% of German e-commerce. The figure rests on reports from 13 networks and platforms and measures sales that passed through affiliate links, not the extra sales created for any one shop [23].

3. Analysis: what AI search and AI assistants change

3.1 Free visitors from Google are falling

The evidence is mounting, although it comes from different markets and methods:

• Pew Research Center analysed the browsing of 900 US adults. On searches that showed an AI summary, users clicked a traditional link in 8% of visits. Without a summary, the figure was 15% [1].
• Ahrefs, a search analytics provider, compared 300,000 keywords. Where an AI overview appears, the average click-through rate of the top result is 58% lower than before. That is a correlation, not proof of cause [2].
• The Reuters Institute surveyed 280 digital leaders from news organisations in 51 countries. They expect search traffic to fall by 43% over three years. The survey is, by its own account, not representative. Data from the analytics firm Chartbeat shows Google traffic to more than 2,500 sites already down by a third between November 2024 and November 2025 [3].
• For UK news sites, Press Gazette reported, using Similarweb data, that the share of searches ending without a click rose from 56% to almost 69% between May 2024 and May 2025 [4].

Our assessment: most of these figures come from the United States and from news publishing. How hard the effect hits online shops in the UK has not yet been measured cleanly. The direction, however, is the same across several independent sources.

3.2 Comparison sites and guide blogs lose reach

People who once compared products on several sites now increasingly ask an AI for the best offer. When AI does that work, the need for pages that only gather information shrinks. For shops, this means that relying on a handful of comparison sites and blogs with search-engine reach carries a growing concentration risk. It makes sense to spread your partners more widely, including channels that do not depend on Google search.

3.3 AI assistants will soon shop for people

The technology for AI-led purchases is arriving at varying speeds:

• In September 2025, OpenAI and the payments company Stripe launched buying directly inside ChatGPT [15]. In March 2026, CNBC reported that OpenAI is shifting its focus to product discovery and letting merchants keep their own checkout [16].
• In May 2026, Google announced a shopping cart that tracks price drops and can complete purchases against criteria the user sets. It starts in the US first, with the UK to follow later [17].
• Amazon is testing "Buy for Me", which starts purchases on brands' own websites, initially for a subset of US customers [18].

Traffic from AI assistants to online shops is growing fast, if from a low base: Adobe reported a 769% rise on the year before for US retail in November 2025 [14]. Shoppers are open to it. Retail Economics and Metapack report that 30% of UK adults are open to AI acting as a personal shopping agent, and that almost half of UK adults under 45 already use AI for product research, price comparison and delivery options (survey of more than 8,000 consumers in eight markets) [12]. For comparison, 21% of German internet users would in principle let an AI buy for them, according to Bitkom [11].

Our assessment: an assistant told to "buy at the best price" checks offers actively. A shop with a valid code is more likely to be chosen than one without.

3.4 Partners with their own community matter more

Telegram channels, newsletters and personal brands speak to their audience directly, without going through search. These are exactly the partners who work in affiliate marketing, and they like nothing better than sharing deals and voucher codes.

3.5 The information stays, the visitor disappears

An AI can summarise a product description. It cannot invent an exclusive discount code. It has to find one somewhere and pass it on. Value therefore shifts: what counts is no longer the visitor on a page but the information that is available somewhere on the web and gets read by AI.

4. Publishers as a source of information for AI systems

4.1 How AI gets its information

A language model only knows what it saw in training. For current questions about offers, prices or reviews, the AI therefore reads web pages while it answers and cites them. Which pages it cites has been studied in some detail:

• Muck Rack analysed more than 25 million links from answers by ChatGPT, Claude and Gemini across 17 industries. Around 27% of cited sources were journalistic articles. The analysis comes from a PR software provider with its own interest in the result [6].
• Profound analysed 680 million citations. Communities are well represented: Reddit accounts for 6.6% of all Perplexity citations, and Wikipedia for 7.8% of ChatGPT's [5].
• The UK agency Glass Digital tested questions in three product categories (kitchen, furniture, cars). Of 479 retrieved editorial pages that could be classified commercially, 420 (87.7%) carried affiliate links. The study states plainly that it does not show whether affiliate partnerships cause a mention. The sample is small and comes from an agency [7].

AI companies are also starting to pay. Press Gazette tracks agreements such as Amazon's deals with Condé Nast and Hearst for its Rufus shopping assistant, and Microsoft's marketplace for publisher content, which pays according to use [9].

4.2 Why publishers are central

Publishers keep producing what AI cannot make itself: current, checked, human-curated information about offers, products and prices. Deal sites, newsletters, creators, communities and guides are, in effect, the raw material for AI answers.

The evidence is clear that AI systems draw heavily on publisher content. From our own day-to-day practice in affiliate marketing we also know that AI assistants pick up voucher codes that publishers post and name them in their answers. That rests on practical experience, not on a study.

4.3 Why that helps the shop

For you as a shop, this changes the sums. Every mention with a publisher is not just advertising space but also a possible recommendation by AI. Publishers are no longer only the place where customers click; they are the source that feeds the AI's buying advice. The advantage is that in affiliate marketing you pay only when a sale happens, not for every impression.

5. The role of voucher codes in the age of AI

5.1 Codes make the shop visible in AI

When someone asks an AI for the best deal, a findable, valid code is a clear signal. It gives the assistant a specific detail it can state in its answer. You appear at exactly the moment the user asks for the best offer, and the code is the reason they buy from you rather than from a competitor. Google, too, is piloting merchant-funded offers in the US that appear in its AI Mode for search [10]. That shows how important offers are becoming at the point of purchase in AI settings.

5.2 Codes work without a click

If an AI buys on the customer's behalf, or the customer arrives through an answer rather than a link, click-based attribution often fails. A dedicated code for each partner still shows who delivered the sale. That also holds where cookies run into limits:

• Google is keeping Chrome's existing choice on third-party cookies and has retired several Privacy Sandbox technologies because of low take-up [19][20]. A full end to cookies in Chrome is therefore off the table, but other browsers apply stricter rules.
• Apple's Safari blocks third-party cookies completely and deletes script-set cookies after seven days without user interaction. For visits arriving through links that carry click identifiers, cookie lifetime is capped at 24 hours [21].
• In the UK, the Data (Use and Access) Act 2025 created a few narrow exemptions from the consent rule for cookies and similar technologies. Going beyond them requires consent, and fines can reach £17.5 million or 4% of worldwide turnover [22].

Our assessment: attribution by code is sturdier because it depends on neither cookies nor a click.

5.3 Codes are easy for AI to read

Discount, code and validity are short, unambiguous details. An AI can recognise and relay them reliably. A US test by Coupons.com shows why quality matters: of the codes ChatGPT, Gemini and Claude gave for 101 retailers, only 24.5% were valid [8]. It is a study by a voucher provider and covers US retailers only. But it shows that AI does not check codes itself. For you, that means a code that is current, clear and permanently active has a better chance of being passed on correctly.

5.4 Deal communities are strong

Deal communities such as HotUKDeals in the UK and mydealz in Germany, and cashback sites such as Quidco and TopCashback, live on offers. They do not depend on Google, and they are places where people go specifically to find a good price. We could not verify current, independently audited user numbers for these services, so we do not quote any.

6. What this means for D2C shops in the UK and Germany

United Kingdom. The Glass Digital study suggests that AI answers in the UK lean heavily on editorial sites with a commercial side [7]. Shoppers are open to AI as a shopping aid [12], and Google's shopping cart is due to reach the UK later [17]. Cookie rules were reset by the Data (Use and Access) Act and remain strict [22]. Cashback sites and deal communities have a long tradition here. Press Gazette's reporting shows that UK publishers are already under pressure from falling search traffic [4].

Germany. For readers who also sell into Germany: the affiliate market is large and established according to BVDW [23], and willingness to use AI shopping aids is high, especially among under-30s [11][13]. Data protection weighs heavily there too. Relying on cookies costs you measurability, while relying on codes keeps it.

Common ground. In both markets, the question of who recommends you matters more and more. How the two markets will diverge over the coming months is open. We offer that as an assessment.

7. Recommendations for shops

1. Give each partner its own exclusive code that stays active permanently. Only a permanent code can establish itself with AI and search engines and be found again and again. A code that expires after two weeks disappears before it has had any effect. With a dedicated code you can see exactly what each partner delivers, even when no click takes place.
2. Spread your partners more widely. Do not rely only on comparison sites and blogs. Add newsletters, creators, communities and deal sites.
3. Win partners with their own community. Telegram channels, newsletters and personal brands are the natural distributors of deals and voucher codes.
4. Measure by code, not only by click. That keeps attribution stable when cookies are missing or an AI triggers the purchase.
5. Pay for results. In affiliate marketing a commission is due only on a real sale. That makes costs predictable while advertising on Google and Meta becomes dearer and less certain.

How selecdoo helps. Through the selecdoo network, your shop and your voucher codes reach newsletters, creators, communities and deal sites, which are the publishers AI draws its answers from. That way you do not depend on a single channel.

8. Conclusion

The internet model in which Google delivers visitors and sites live off them is losing strength, because AI answers replace clicks. At the same time, the need for accurate, current information remains. Publishers supply it, and so become one of the most important sources for AI systems. That is an opportunity for shops: if you are represented with publishers through exclusive, permanent codes, you become visible in answers and can win the sale.

The question is no longer "How do I rank on Google?" but "Who recommends me when people and AI look for the best offer?"

Sources

1. Pew Research Center (22 July 2025): Google users are less likely to click on links when an AI summary appears in the results
2. Ahrefs (4 February 2026): AI Overviews Reduce Clicks (Update)
3. Reuters Institute (12 January 2026): Journalism, media, and technology trends and predictions 2026
4. Press Gazette (7 July 2025): How Google AI Overviews is fuelling zero-click searches for top publishers
5. Profound (5 June 2025): AI Platform Citation Patterns
6. Muck Rack (as updated 7 May 2026): What Is AI Reading? (May 2026)
7. Glass Digital (2 September 2026): Do affiliate mentions help AI visibility? We tested three UK categories
8. Coupons.com (16 September 2026), press release via Yahoo Finance: Coupons.com study reveals AI chatbots and discount codes
9. Press Gazette (as updated 2 November 2026): News publisher AI deals and lawsuits
10. Google (11 January 2026): Agentic commerce AI tools and protocol for retailers and platforms
11. Bitkom (24 November 2025, in German): Black Friday: Hälfte auf Schnäppchenjagd
12. Retail Economics and Metapack, reported by ecommercenews.uk (5 February 2026): AI drives UK retail as shoppers embrace ecommerce tools
13. Criteo, reported by MEEDIA (19 February 2026, in German): Zwei von drei Deutschen wollen 2026 mit KI einkaufen
14. Digital Commerce 360 (13 January 2026), citing Adobe Analytics: Generative AI and online holiday shopping traffic 2025
15. Stripe (29 September 2025): Stripe powers Instant Checkout in ChatGPT
16. CNBC (24 March 2026): OpenAI revamps shopping experience in ChatGPT after Instant Checkout
17. Google (19 May 2026): Universal Cart: your new hub for shopping on Google
18. Amazon: Buy for Me
19. Google Privacy Sandbox (22 April 2025): Next steps for Privacy Sandbox and tracking protections in Chrome
20. Google Privacy Sandbox (17 October 2025): Update on plans for Privacy Sandbox technologies
21. WebKit: Tracking Prevention in WebKit
22. Stevens & Bolton LLP (6 May 2026): The Data (Use and Access) Act 2025: ICO guidance on cookies and the use of storage
23. BVDW (17 June 2026, in German): APMC-Studie macht Marktpotenzial von Affiliate- und Partner-Marketing sichtbar

Note: all sources were retrieved on 3 November 2026. US figures are labelled as such in the text.