Hyper-personalization is the practice of tailoring an advertisement, offer or piece of content to a single individual at the moment it is shown, using live behavioural and contextual signals and machine learning that predicts what that person is likely to respond to. Ordinary personalisation chooses which ad a person sees from data about them. Hyper-personalisation goes further: it changes the content itself, and increasingly generates it, for each impression. It exists because marketers believe relevance drives response, and AI has made thousands of creative variants cheap to produce.

The term has no standard definition. As of October 2026 it does not appear in any IAB Tech Lab specification, and no regulator or platform policy uses it as a defined category. It is a vendor and consultancy label. IBM, in a January 31, 2025 explainer, separates the two ideas on two axes: traditional personalisation uses names, purchase history and demographics, and reacts to the past; hyper-personalisation adds browsing activity, location, device, time of day and weather, and uses predictive analytics to anticipate needs. Deloitte Digital, in a June 2026 overview, calls it "the most advanced form of personalized customer engagement". Both describe a direction of travel, not a threshold.

How it works

Most systems sold under the label have three layers.

The data layer assembles what is known about the person. That means first-party records such as purchases and loyalty membership, behavioural signals such as recent clicks and searches, and contextual signals such as location, device, time and weather. Getting that data into the systems that buy and render ads is the job PPC Land has described as data activation; where the advertiser buys records about people from companies with no direct relationship to them, it is third-party data.

The decision layer predicts. A model scores the person against outcomes such as purchase, churn or price sensitivity, and estimates which product, message, discount or format will perform best. The question is not only whether this person should see an ad, but what the ad should say to them now.

The content layer renders the answer. In the older model, a designer built a template with swappable slots for headline, image, price and call to action, and a rules engine filled them. In the newer model, generative AI writes the copy, edits the image or assembles a video at the time of the impression.

Amazon's Dynamic TV Creative, announced on May 11, 2026, is a worked example. An advertiser uploads one base creative for Prime Video. At impression time, Amazon matches the viewer to a likely stage in the purchase journey using shopping, browsing and streaming signals tied to logged-in accounts, then changes the product imagery, the headline, the product details shown (ratings, Prime benefits or price) and the call to action, which can read "Add to Cart", "Send to Phone", "Save to Cart" or "Visit Brand Store". Amazon reported that dynamic creative across its DSP delivered 7.2 per cent higher return on ad spend (ROAS) in internal data from August 2025. It had not published results for the TV product itself.

Walled gardens such as Meta, Google and Amazon run the whole loop internally, because they hold logged-in identity, inventory and creative tools in one place. In the open programmatic market, the pieces are split between a customer data platform, a demand-side platform (DSP) and a creative vendor.

Origin and evolution

The building blocks are older than the label. Dynamic creative optimisation (DCO), which assembles display ads from components according to data about the viewer, was commercial by the late 2000s. Google announced its purchase of the display creative company Teracent on November 23, 2009.

The label itself has no recorded coinage. It spread through vendor and consultancy material, while the reputation of personalisation swung. On December 2, 2019, Gartner predicted that by 2025, 80 per cent of marketers who had invested in personalisation would abandon the effort, citing weak returns and the burden of managing customer data. "Consumers have developed an increasingly jaundiced eye toward marketers' efforts to embrace them," said Charles Golvin, a senior director analyst at the firm. Two years later, on November 12, 2021, McKinsey reported that 71 per cent of consumers expected personalised interactions and 76 per cent were frustrated when they did not get them, and that personalisation most often lifted revenue by 10 to 15 per cent.

Generative AI changed the economics. Template-based DCO could only recombine what a designer had made. Large language and image models can produce new variants without limit. Mark Zuckerberg described this as "infinite creative" in a May 1, 2025 interview with Stratechery, in which he set out a model where a business states its objective and connects a bank account, and Meta handles the rest: "We just make it for them." Three weeks later, at Google Marketing Live on May 21, 2025, Google said AI Max for Search would generate creative and landing pages based on user context. Aravind Srinivas, chief executive of Perplexity, spoke on a podcast that same month of using browsing context to build user profiles for "hyper personalized" advertising.

Why it matters to marketers

For advertisers on the large platforms, hyper-personalisation is less a strategy to adopt than a default to manage. Meta made dynamic media the default for Advantage+ catalog ads from September 1, 2025, with full enforcement on October 20, 2025, according to PPC Land's Advantage+ explainer. Meta began using conversations with its AI assistant to personalise ads on December 16, 2025, excluding sensitive topics such as health, religion and sexual orientation. OpenAI's ads system can draw on past chats and memory when users enable personalised ads.

Performance claims are mostly self-reported. Meta said in June 2025 that Advantage+ sales campaigns lifted ROAS by an average of 22 per cent, figures it presented when it announced new generative tools at Cannes Lions.

The bottleneck has moved from data to content. Toby Coulthard of Jacquard argued in December 2025 that hyper-segmentation creates unsustainable content demands, which is precisely the gap generative tools are sold to fill.

Limits, law and disputes

Evidence. One incrementality test found that Advantage+ produced about 17 per cent of the conversions Meta's own attribution claimed. Gartner found in June 2025, from a survey of 1,464 buyers fielded in late 2024, that 53 per cent of customers had negative experiences of personalised marketing, and that those customers were 3.2 times more likely to regret a purchase. "Passive personalization tactics alone no longer suffice," said Audrey Brosnan, a senior director analyst.

Brand control. When the platform writes or edits the ad, the advertiser may not see what runs. In June 2026 REI pulled an Instagram ad whose bicycle photo had been altered into a machine with extra handlebars, saying Meta had "auto-enrolled us in an AI personalization tool".

Privacy law. In Europe, the General Data Protection Regulation (GDPR) defines profiling in Article 4(4) as automated evaluation of personal aspects such as preferences, behaviour and location, and Article 21(3) gives people an absolute right to object to profiling for direct marketing. The ePrivacy Directive requires consent before reading from or writing to a user's device. In October 2024 the Court of Justice of the EU ruled that personalised advertising cannot use unlimited quantities or types of personal data, a direct constraint on the "more signals" logic of the practice. In the US, comprehensive state privacy laws, beginning with Virginia's in 2023, give residents the right to opt out of targeted advertising and certain profiling.

Manipulation and price. The EU AI Act bans AI techniques that materially distort behaviour through manipulation, with prohibitions applying from February 2, 2025. Where tailored persuasion ends and manipulation begins is contested; the researcher Theo Antunes summarised the line as "influence is permissible, but manipulation is strictly prohibited". The sharpest dispute concerns prices rather than messages. The US Federal Trade Commission ordered eight companies, including Mastercard, Accenture and McKinsey, to explain their surveillance pricing practices on July 23, 2024, and on August 19, 2026 proposed a policy statement treating undisclosed personalised pricing as a likely violation of the FTC Act. "When consumers see a listed price, they expect it to be same price that everyone else sees," said Chairman Andrew Ferguson.

Not the same as

Ad personalisation. The broader practice of selecting which ad a person sees based on data about them, covered in PPC Land's ad personalization explainer. Hyper-personalisation is a subset that also changes the ad's content.

Dynamic creative optimisation. A specific technology that assembles ads from pre-built components. DCO is one way to deliver hyper-personalisation; generative AI is another.

Segment-based targeting. Showing one message to a group defined by shared traits. Lookalike modeling expands such groups. Hyper-personalisation claims to address the individual rather than the segment, though in practice many systems still operate on fine-grained segments.

Personalised pricing. Changing the price, not the message, for each customer. It uses the same data and models but faces separate consumer protection rules.

Recent developments

At Cannes Lions on June 23, 2026, Meta said it was testing an end-to-end creative system that analyses campaign results and generates new ads from them, and that more than 8 million advertisers used its generative creative tools. A separate update the same month said Sales campaigns would use product data to assemble the best-performing ad for each viewer in real time. Amazon plans to expand Dynamic TV Creative to live sports and Prime Video Channels in the third quarter of 2026.

Timeline

  • November 23, 2009 - Google announces acquisition of Teracent, a dynamic display creative company.
  • December 2, 2019 - Gartner predicts 80 per cent of marketers investing in personalisation will abandon it by 2025.
  • November 12, 2021 - McKinsey reports 71 per cent of consumers expect personalised interactions.
  • January 1, 2023 - Virginia's Consumer Data Protection Act takes effect, with opt-outs for targeted advertising and profiling.
  • July 23, 2024 - FTC issues 6(b) orders on surveillance pricing to eight companies.
  • August 1, 2024 - EU AI Act enters into force.
  • October 2024 - CJEU rules personalised advertising cannot use unlimited personal data.
  • January 31, 2025 - IBM publishes its hyper-personalisation definition.
  • February 2, 2025 - AI Act prohibitions on manipulative AI practices apply.
  • May 1, 2025 - Mark Zuckerberg describes "infinite creative" in a Stratechery interview.
  • May 21, 2025 - Google announces AI Max for Search with generated creative and landing pages.
  • June 3, 2025 - Gartner reports 53 per cent of customers had negative experiences of personalisation.
  • June 17, 2025 - Meta announces generative AI advertising tools at Cannes Lions.
  • September 1, 2025 - Dynamic media becomes the default for Meta Advantage+ catalog ads.
  • December 16, 2025 - Meta begins using AI assistant conversations for ad personalisation.
  • May 11, 2026 - Amazon announces Dynamic TV Creative for Prime Video.
  • June 22, 2026 - REI removes an Instagram ad altered by a Meta AI tool.
  • June 23, 2026 - Meta previews an end-to-end AI creative system at Cannes Lions.
  • August 19, 2026 - FTC proposes a personalised pricing enforcement policy statement.

Summary

Who. Advertising platforms including Meta, Google, Amazon and OpenAI run hyper-personalisation inside their own systems; customer data platforms, DSPs and creative vendors supply it in the open market. Consultancies such as Deloitte, IBM and McKinsey promote the term. Regulators including the European Commission, EU courts, the FTC and US states set its limits.

What. It is the tailoring of ad content, offers and sometimes prices to one person at the moment of exposure, using live behavioural and contextual data, predictive models and increasingly generative AI. It has no standard definition and functions as a vendor label.

When. Dynamic creative technology was commercial by 2009. The strategy term spread through consulting in the 2010s, and generative AI tools from Meta, Google and Amazon in 2025 and 2026 made per-impression creative a platform default.

Where. Mostly inside logged-in walled gardens, including social feeds, search, retail media, streaming TV and AI chat assistants, and in programmatic display and video through DCO vendors.

Why. Marketers expect relevance to raise response, and AI has cut the cost of producing variants. The evidence for incremental gains is largely self-reported, and the practice sits close to legal lines on profiling, data minimisation, manipulation and price discrimination.