Preparing for an Era of Deepfakes and AI-Generated Ads: A Framework for Understanding Responses to Manipulated Advertising

Part of: Digital Marketing Topic 8 — Trends Continued: Deepfakes & Manipulated Advertising · Reading Citation: Campbell, C., Plangger, K., Sands, S., & Kietzmann, J. (2022). Preparing for an era of deepfakes and AI-generated ads: A framework for understanding responses to manipulated advertising. Journal of Advertising, 51(1), 22–38. Key concepts: Synthetic Advertising, Deepfakes, Generative Adversarial Networks, Ad Manipulation, Verisimilitude, Persuasion Knowledge

---

TL;DR

Advertising is shifting from analog and digital editing toward synthetic advertising — ads generated or altered automatically by AI (deepfakes, GANs) that depict a convincing but artificial reality. The authors build a conceptual framework explaining how consumers respond to all forms of ad manipulation along a spectrum of "manipulation sophistication". Greater sophistication raises perceived verisimilitude (how real an ad looks) and perceived creativity, both of which boost persuasion but can also trigger awareness of ad falsity, which dampens it. The effects intensify when personal data is used to hyperpersonalise ads. The paper closes with a research agenda built around three areas: ad falsity, consumer response, and originality.

Why It's on the Reading List

It is the anchor reading for the deepfakes/manipulated-advertising trend: it provides the canonical typology (analog → digital → synthetic) and the eight-proposition framework that the exam is likely to test on how consumers react to AI-generated ads and the privacy/authenticity trade-offs they create.

Background & Research Question

Manipulation is "as old as advertising itself" — advertisers have always altered data to make ads appealing. What has changed is the sophistication of that manipulation. The paper argues research on analog and digital manipulation has been fragmented and ad hoc, leaving no unified theory for the newest, most advanced form: synthetic advertising. The aim is therefore conceptual — to (1) characterise manipulation along a spectrum and (2) build a comprehensive framework explaining differential consumer responses, grounded in existing advertising theory.

Key Concepts & Definitions

Definition — Ad manipulation

Any technique used to alter an ad — across preproduction (clothing, makeup), production (lights, lenses) and/or postproduction (retouching). All ads are therefore "artificial representations of reality".

Definition — Synthetic advertising

Ads generated or edited through the artificial and automatic production and modification of data, typically using AI algorithms (deepfakes, GANs) to depict a convincing yet fake version of reality. The most advanced form of manipulated advertising to date.

Definition — Verisimilitude

"A likeness to truth" — the extent to which a consumer perceives an ad to be true or real. Easier to achieve with more sophisticated techniques.

Definition — Awareness of ad falsity

The extent to which a consumer perceives an ad to be false. Deliberately placed at the centre of the framework. Covers three falsity types: false product-related claims, product-unrelated claims, and false presented reality.

Three generations of manipulation

Not successive — evolutionary

These are not generations where one replaces another; they coexist and are often combined creatively.

Main Arguments / Findings

How the technologies work.

• Deepfakes swap the attributes (face, voice, skin tone, gender) of a source onto a target by training a deep neural network called an autoencoder. The encoder extracts ~300 dimensions (abstract facial characteristics, expression) into a compressed "latent space"; the decoder reconstructs the input. It is a "lossy" process but can generate expressions the person never actually conveyed. (Example: Charlize Theron's Dior J'adore ad re-faced with Rowan Atkinson as Mr Bean.)
• Generative Adversarial Networks (GANs) generate original synthetic media using two oppositional networks: a generator creates synthetic output, a discriminator tries to distinguish it from real data. They iterate until output is indistinguishable — enabling entirely fabricated models, clothes, portraits, age/gender swaps, and text-to-image generation.

Synthetic advertising is already accessible

Start-up Rosebud AI lets agencies instantly alter a model's ethnicity, age, expression or gender. Deepfakes translated a speech by Indian politician Manoj Tiwari into Hindi + 20 dialects. GANs can produce synthetic models with no human shoot.

Manipulation and personalisation. Synthetic tools can reach hyperpersonalisation — tailoring content in real time per customer using social-media, sensor or loyalty data (imagine an ad with a model matching your ethnicity, wearing clothes you've bought, on a street near home). But this heightens awareness of customer surveillance, activating privacy concerns and feelings of vulnerability — especially when data collection is covert.

The double edge

The same sophistication that makes ads more persuasive (via verisimilitude and creativity) can also cue consumers that an ad is fake — and personalised data use can make falsity "jarringly obvious", triggering reactance.

Framework / Model

The framework links manipulation sophistication (the polish/finesse of content creation) to persuasion outcomes through three mechanisms, formalised as eight propositions.

Key logic:

• High verisimilitude lets consumers process an ad as if real, enabling persuasive processes — mental imagery and narrative transportation — and blocks falsity awareness (consumers avoid/discount ads they know are manipulated).
• Creativity = originality + relevance. It boosts persuasion (attention, recall, liking) AND can trigger falsity awareness (fictional/mythological content, or deeply personalised content, cues "this is fake").
• Falsity awareness reduces persuasion via three routes: persuasion knowledge (Friestad & Wright), preference for authenticity over fakeness, and inferences about advertiser investment/cost.
• But P8: creativity can make consumers condone a manipulated ad if its value (entertainment, relevance, time-saving) is high enough — an "advertising calculus" weighing benefits against costs.

Advertising calculus

Analogous to the privacy calculus — consumers weigh an ad's benefits (personalisation, relevance) against its costs (privacy, surveillance) in deciding how to react.

Implications for Marketers

• Synthetic tools democratise creativity — costly creative ideas (exotic locales, gravity-defying stunts, deceased celebrities) become financially feasible, levelling the field and aiding D2C challengers.
• This makes it harder to signal high ad spend, threatening luxury brands that use production value for differentiation.
• Personalisation pays off but risks reactance, privacy concerns and vulnerability — especially with covert surveillance; balance benefits against perceived costs.
• Expect future regulation/disclosures (e.g. Photoshop disclaimers in beauty), and growing "manipulated advertising literacy" as consumers gain experience detecting and tolerating synthetic ads.

Exam Takeaways

Likely exam points

• The three generations: analog → digital → synthetic (know examples and what defines each).
• Deepfakes (autoencoders, edit existing media) vs GANs (generator/discriminator, generate new media).
• The framework's three mechanisms: verisimilitude, creativity, awareness of falsity — and the eight propositions' signs.
• Falsity awareness sits at the centre; it is reduced by verisimilitude (P4) but raised by creativity (P6).
• Hyperpersonalisation + covert customer surveillance → privacy concerns, vulnerability, reactance ("advertising calculus").
• Three future-research areas: ad falsity, consumer response, originality.

Summary

• Conceptual paper (no empirical study); a literature review feeding a framework + research agenda.
• All advertising is manipulated to some degree; synthetic advertising is the most advanced form.
• More sophistication → more verisimilitude and creativity → more persuasion, but also potential falsity awareness that undercuts it.
• Personal data amplifies both the upside (relevance) and the downside (privacy/vulnerability).

Related Notes

• Digital Marketing — subject hub
• Synthetic Advertising — the central phenomenon
• Deepfakes · Generative Adversarial Networks — the enabling AI techniques
• Persuasion Knowledge — why falsity awareness reduces persuasion
• Narrative Transportation — a persuasive process enabled by high verisimilitude
• Customer Surveillance — the privacy cost of hyperpersonalisation