Chapter 13: Ethics, Responsible AI and Sustainability

Dana Okafor has three proposals on her desk, and each one could make Lumen Cycles more money. Marcus Lee wants to use the ride data from the Lumen app to send personalized offers: a rider who commutes 12 miles a day might get a battery-upgrade offer, and a rider whose app shows trips to a school twice a day might get an ad for the Lumen Cargo. Jess Patel wants to replace expensive photo shoots with AI-generated lifestyle images of smiling families on Cargo bikes. And a pricing consultant has pitched dynamic pricing on the website, with prices that move with demand, inventory and each visitor's browsing behavior.

None of these ideas is illegal on its face. All of them are things real companies do. Yet each one made Dana uneasy, and she couldn't say exactly why. Would riders feel watched if an ad seemed to know where they took their kids every morning? Would a parent feel deceived on learning that the "real Lumen families" in an ad never existed? Would two neighbors be upset to find they'd been charged different prices for the same bike on the same day?

This chapter gives you the tools to answer questions like Dana's: the long-standing criticisms of marketing, the expectations around customer data, the specific risks AI adds, and sustainable marketing. Earlier chapters raised these issues one at a time. Here they come together, with a practical checklist for any marketing decision that involves AI.

13.1 Criticisms of marketing

Marketing has critics, and some of the criticism is earned. Understanding the critics' arguments helps you see real ways marketing goes wrong, and predict where regulators, journalists and customers will push back.

Law and ethics overlap, but they aren't the same. Plenty of legal practices are unethical (writing a cancellation process so confusing that customers give up), and the law often trails behind new technology. A useful test is whether you'd be comfortable if your customers, your family and a reporter saw exactly what you did and why. If the honest answer is no, the practice deserves a second look even if a lawyer has approved it.

Critics' arguments fall into three groups: marketing's effects on individual consumers, on society as a whole, and on other businesses.

13.1.1 Effects on individual consumers

High prices. Critics argue that marketing raises prices through the costs of distribution, advertising and promotion, and through excessive markups. Marketers respond that intermediaries do real work (carrying inventory, offering test rides, providing service), that advertising helps products reach the scale that lowers unit costs, and that competition limits markups. Both sides have a point.

Deceptive practices. These include misleading prices (a "sale" price compared with a "regular" price that was never really charged), misleading promotions (contests with hidden conditions), and misleading packaging or claims (a box that is mostly air, or a "fast-charging" claim that holds only in ideal conditions). In the United States the Federal Trade Commission (FTC) polices deceptive practices, and other countries have their own consumer-protection agencies. The harder cases involve puffery, the vague exaggeration ("the best ride of your life") that the law usually tolerates because no reasonable buyer takes it literally.

High-pressure selling. Salespeople and websites can push people into purchases they wouldn't otherwise make. Online, high pressure often takes the form of dark patterns: interface designs that steer users into choices they didn't intend, such as countdown timers that reset, pre-checked boxes for add-ons, "confirmshaming" buttons ("No thanks, I don't care about my family's safety"), or subscriptions that take one click to start and a phone call to cancel. Regulators in the United States and Europe have increasingly treated these designs as deceptive or unfair, and have taken action against companies over hard-to-cancel subscriptions.

Shoddy, harmful or unsafe products. Some products don't perform as promised, and some are dangerous. E-bikes offer a current example: news about lithium-ion battery fires, often involving cheap or uncertified batteries, has made battery safety a public concern and prompted rules about battery standards in some cities.

Planned obsolescence. Critics charge that some companies design products to wear out or become outdated sooner than necessary, through parts that can't be replaced or software support that ends early. Sometimes newer models simply are better. The "right to repair" movement, which has won new laws in several places, pushes companies to make parts, tools and manuals available so products last longer.

Poor service to disadvantaged consumers. Lower-income and marginalized communities can face fewer stores, higher prices, worse credit terms and more aggressive selling. The issue gets sharper when targeting is automated, as you'll see in the section on responsible AI. A system that learns to avoid "low-value" neighborhoods can reproduce old patterns of exclusion without anyone deciding to.

13.1.2 Effects on society

False wants and too much materialism. Critics argue that marketing teaches people to define themselves by what they own. Marketers respond that people have always wanted status and novelty. The truth is in between: marketing can't make people want things they have no use for, but it can shape which wants feel normal and urgent.

Too few social goods and cultural pollution. Private goods often carry public costs, which economists call externalities: every car sold adds traffic, pollution and demand for roads the buyer doesn't fully pay for. Critics also object to the sheer volume of commercial messages in feeds, search results and public spaces. Ad blockers and paid ad-free subscriptions are partly a consumer response.

13.1.3 Effects on other businesses

Critics also charge that some practices hurt competition: buying up rivals, raising barriers through exclusive deals with distributors, or pricing below cost to drive a competitor out. Competition law addresses these practices, and the questions have new urgency now that a few large platforms control the search engines, app stores and marketplaces smaller businesses depend on.

13.1.4 How society and companies respond

Two citizen movements have pushed marketing to change. Consumerism is the organized movement of citizens and government agencies to improve the rights and power of buyers: the right to safe products, to accurate information, to choose, and to be heard. Environmentalism is the organized movement of people concerned with protecting the natural world and the quality of life. Both have led to regulation, and both have changed what customers expect from the brands they buy.

The most important response, though, is the one companies control: choosing not to do the thing the critics describe. That choice is easier when a company sees ethics as part of long-term customer value rather than a cost. A customer who feels tricked doesn't come back, and in a world of online reviews, they tell a lot of other people. The customer lifetime value idea from Chapter 6. Marketing Research and Analytics applies here: a practice that adds a sale today but loses a customer for life is usually a bad trade, even before you count the damage to the brand.

13.2 Privacy and data

Modern marketing runs on data. Every click, purchase, location ping and app session can become an input to segmentation, targeting, personalization and measurement. That data makes marketing more relevant and less wasteful. It also gives companies an unprecedented view into people's lives, and customers know it.

13.2.1 Why privacy is a marketing issue

It's tempting to treat privacy as a legal department problem. That's a mistake for two reasons.

First, privacy is about trust, and trust is a marketing asset. Customers share more data with companies they trust when they see a fair exchange: a store remembers their size, an app reminds them when service is due. When a company uses data in a way customers didn't expect, the reaction can be strong. A widely retold story about a large US retailer whose purchase model reportedly inferred a shopper's pregnancy, and sent baby-product coupons, spread because it captured a common fear: that companies know more than people have chosen to reveal.

Second, privacy failures can be enormous. The Cambridge Analytica scandal, which became public in 2018, involved data on millions of Facebook users collected through a quiz app and later used for political targeting. It led to investigations, large penalties and lasting damage to Facebook's reputation.

13.2.2 Consent and the data relationship

The central idea in data ethics is consent: people should know what data is collected about them and have a real choice about how it's used.

Consent is easy to fake. A 6,000-word privacy policy that nobody reads, accepted with a single "I agree" button, satisfies the letter of consent but not its spirit. Responsible marketers aim for consent that is:

• Specific. It covers a particular use. Agreeing to service reminders is not agreeing to targeted advertising.
• Informed. It is explained in plain language at the moment it matters, not buried in legal text.
• Freely given. People can say no without losing the core service.
• Reversible. People can change their minds and withdraw it easily.

Marketers also distinguish opt-in consent, where people must actively agree before data is used, from opt-out consent, where the use happens by default unless people object. Opt-in respects choice more, and it usually produces smaller but more willing audiences.

A related principle is data minimization: collect only the data you need for a clear purpose, keep it only as long as you need it, and don't repurpose it without asking. Data you never collected can't be breached, misused or subpoenaed.

When Apple introduced App Tracking Transparency in 2021, apps on its devices had to ask permission before tracking users across other companies' apps and websites. Many users said no, and advertisers who depended on that tracking had to rethink their targeting. The lesson: given a clear, easy choice, many customers decline to be tracked. A data strategy that depends on customers not noticing is fragile, which is one reason first-party data, collected directly from a company's own customers, has become more valuable.

13.2.3 Data protection rules

Privacy law differs from place to place and changes often, so this section describes it in general terms. You should always check current rules for the markets you work in, and work with legal experts on the details.

• Comprehensive data protection laws. The European Union's General Data Protection Regulation (GDPR), in effect since 2018, is the best-known example. Laws like it generally require a lawful basis for using personal data, give people rights to access, correct and delete it, and impose substantial fines. Many countries have similar laws.
• US state privacy laws. The United States has no single comprehensive federal privacy law. California and a growing number of other states give residents rights to know what data is collected, to delete it and to opt out of its sale or use for targeted advertising.
• Sector and sensitive-data rules. Health, financial and children's data are covered by stricter laws, and many laws give extra protection to precise location, race or ethnicity, religion, sexual orientation and biometric data, the categories where misuse does the most harm.
• Consumer-protection law. Regulators can act when a company's data practices contradict its own privacy promises, because breaking such a promise is a deceptive practice.

13.2.4 Data security

Privacy also means protecting the data you hold. Breaches expose customers to fraud and lead to penalties, lawsuits and lost trust. Marketing teams use many outside tools, from email platforms to AI assistants, and each one that touches customer data is another place it can leak.

This connects directly to the protect data rule from Chapter 1. Marketing in an AI Era. Pasting a customer list or confidential sales figures into a public AI tool can send that data to a company whose terms let it be stored or used in ways you can't control. Many companies now provide approved AI tools with contractual data protections and forbid unapproved ones for customer or confidential data.

Companies that handle data well share a few habits: they explain the value exchange in plain language, give customers real controls, use data only for purposes customers agreed to, avoid sensitive inferences (a health condition, a pregnancy, financial trouble) without explicit consent, and delete what they no longer need.

13.3 Responsible AI in marketing

AI makes almost every marketing activity in this book faster and cheaper: writing copy, generating images, segmenting customers, setting prices, answering questions and placing ads. It also makes some old ethical problems bigger and creates a few new ones. Each earlier chapter raised a piece of this. Here they come together.

Responsible AI is not anti-AI. A company that refuses to use AI at all may serve its customers worse, with slower service and less relevant offers, and fall behind competitors that use AI well. The goal is to get AI's benefits without harming customers, misleading them or losing their trust. Five risks matter most for marketers.

13.3.1 Bias

AI systems learn patterns from data, and data reflects the world as it has been, including its unfairness. If past decisions favored some groups, a system trained on those decisions will tend to favor them too.

A well-known example comes from hiring rather than marketing. Amazon built an experimental tool to rate job applicants, trained on résumés the company had received over the previous decade, most of which came from men. Reuters reported in 2018 that the tool learned to downgrade résumés that signaled the applicant was a woman, and Amazon abandoned it. Nobody programmed the bias in. The system learned it from history.

Marketing has its own versions. Ad-delivery systems that optimize for clicks or conversions can end up showing ads mostly to some demographic groups and not others, even when the advertiser didn't ask for that. When the ads are for housing, jobs or credit, such patterns can raise legal problems under anti-discrimination law. Facebook's ad system was the subject of a US government housing-discrimination case over exactly this issue, which the company settled by agreeing to change how its system delivers housing ads.

Bias also hides in proxies. A model that never sees race or income may still use zip code, device type or shopping times, which correlate with them. Dynamic pricing that charges more in neighborhoods with fewer competitors can end up charging more to lower-income or minority neighborhoods without anyone intending it. Chapter 9. Pricing raised this risk, and Chapter 7. Segmentation, Targeting and Positioning raised the parallel risk for micro-segments.

Generative AI adds another kind of bias: representational bias. Ask an image generator for "a CEO" or "a nurse" or "a family riding bikes," and the results may lean heavily toward certain ages, races, genders and body types. A brand that publishes those images without review can quietly narrow who appears to belong in its world.

The practical response is to test outcomes, not just inputs. Look at who sees an ad, who gets an offer, what price each group pays and who appears in generated images, broken down by groups that matter. If the differences can't be justified by a legitimate business reason that doesn't depend on a protected characteristic, change the system.

13.3.2 Manipulation

Persuasion is a normal part of marketing. Manipulation is persuasion that works by getting around a person's reasoning rather than through it: exploiting a weakness, a bias or an emotional state the person wouldn't endorse if they saw it clearly.

AI raises the stakes in two ways. First, scale and precision: a system that tests thousands of message variations can find which framing works on which person, including framings that exploit anxiety or impulsiveness. Chapter 4. Consumer Behavior discussed how persuasion based on personal data can cross this line. Second, conversation: an AI assistant built to maximize sales could learn to flatter, pressure or invent urgency one on one, in ways no human salesperson would get away with.

A few questions help separate persuasion from manipulation:

• Would the customer still make this choice if they understood exactly how the message was designed and targeted?
• Does the tactic rely on a weakness, such as a known gambling problem, a late-night impulse or grief, rather than on the product's merits?
• Are vulnerable groups, such as children, older adults or people in financial distress, more exposed to it?
• Is important information (total price, cancellation terms, risks) easy to find?

13.3.3 Deepfakes and synthetic content

Generative AI can produce realistic images, video, voices and text that never came from a real person or event. Many uses are harmless and clearly fictional, such as an animated mascot or a stylized illustration. The problems start when synthetic content is presented as real.

Marketers face several versions of this risk:

• Fake people presented as real: invented "customers" in testimonials, "real families" in lifestyle ads, or generated influencers. Chapter 12. Digital, Social and Search Marketing discussed AI influencers and disclosure.
• Fake reviews. AI makes plausible reviews cheap to write. They're deceptive whoever writes them, and US regulators have moved to ban them explicitly, including AI-generated ones.
• Unauthorized likenesses. Using a celebrity's or employee's voice or face without permission can violate their rights and imply a false endorsement.
• Brand impersonation. In a widely reported 2024 case, an employee of an engineering firm transferred company funds after a video call in which the other "executives" were deepfakes. Brands increasingly monitor for impersonation and tell customers how to recognize real communications.

The line to hold is simple: don't present synthetic content as real when the difference matters to the audience. An illustrated bike in a whimsical ad needs no label. A "real customer story" must be a real customer's story.

13.3.4 Transparency

People generally want to know when they're dealing with a machine, when content was generated, and why they're seeing a particular ad, price or recommendation.

Transparency operates at several levels:

Level	Question customers ask	Responsible practice
Interaction	"Am I talking to a person or a bot?"	Label AI chat assistants clearly, and offer a route to a human
Content	"Is this image, review or person real?"	Disclose AI-generated or AI-altered content when it could mislead
Decision	"Why did I get this price, ad or offer?"	Give plain-language reasons and an easy way to adjust preferences
Data	"What did you use to decide that?"	Explain which kinds of data drive personalization, and let people opt out

Rules are moving in this direction. The European Union's AI Act, adopted in 2024, includes transparency requirements such as telling people when they're interacting with an AI system and labeling certain AI-generated content. Several US states require disclosure of chatbots or AI-generated content in particular settings, such as political ads, and advertising platforms have their own policies. The details change quickly, but the trend is clear: treat disclosure as the default.

This extends the disclose it rule from Chapter 1. Marketing in an AI Era. In your coursework, you disclose AI's role to your instructor. In marketing practice, a company discloses AI's role to its customers.

13.3.5 Accountability

When an AI system makes a mistake, someone has to answer for it. Customers, courts and regulators don't accept "the algorithm did it" as an excuse.

A widely cited case shows why. In 2024 a Canadian tribunal ruled against Air Canada after the airline's website chatbot gave a customer wrong information about a bereavement fare. The airline argued, in effect, that the chatbot was responsible for its own statements. The tribunal disagreed: the chatbot was part of the airline's website, and the airline was responsible for everything on it. For marketers the lesson is direct. Whatever your AI tells a customer, your company told the customer.

Accountability in practice means:

• A named owner for each AI system or AI-assisted process, with authority to pause it.
• Human oversight proportional to the risk. A human reviews every published ad. A human handles every warranty claim. A pricing algorithm runs within limits a person has set and reviews regularly.
• Monitoring for errors, complaints and unfair outcomes after launch, not just testing before.
• Recourse for customers: a way to reach a person, challenge a decision and get it fixed.
• Records of what the system does and how it was checked, so that decisions can be explained later.

This is the own the decision rule from Chapter 1. Marketing in an AI Era, applied to a whole organization.

13.3.6 What AI does well, and where judgment stays human

AI doesn't only create ethical risk. It can also help marketers act responsibly: scanning ads for unsubstantiated claims, flagging biased language, checking that campaign images show a range of customers, monitoring for impersonation scams and summarizing complaints so patterns of harm show up sooner.

What AI can't do is decide what is fair, what customers deserve to know, or which risks are worth taking. Those are value judgments, and they belong to people who can be held responsible for them. An AI tool can tell you that an ad reaches 70% men; it can't tell you whether that's a problem for your product and your customers.

13.3.7 The TRUST check

The five risks above can be turned into a practical checklist you run before any AI-assisted marketing decision goes live. It extends the four rules from Chapter 1. Marketing in an AI Era from your own work to your customers, and adds one new question about fairness. Call it the TRUST check.

Letter	Question to ask	Warning signs	Connects to the rule
Tell people	Would customers know, or be comfortable learning, that AI was involved here? Is AI's role disclosed where it matters?	Bots posing as people; synthetic images presented as real customers; a reluctance to explain how it works	Disclose it
Respect their data	Was the data collected with consent for this use? Is it the minimum needed? Is any of it sensitive?	Data repurposed from another purpose; precise location, health or children's data; customer data pasted into public tools	Protect data
Unbiased outcomes	Have we tested who is included, excluded, charged more or shown less, across groups that matter?	Results that differ sharply by area, age, gender or race; proxies like zip code; generated images that show only one kind of person	(new)
Straight dealing	Are all claims true and verified? Does the tactic persuade with accurate reasons rather than exploit weaknesses?	Unverified AI-written claims; fake urgency; fake reviews; targeting people at vulnerable moments	Verify it
Take ownership	Who is the named owner? Can a human review, override and explain the result? Can customers reach a person and get errors fixed?	"The system decided"; no monitoring after launch; no human route for complaints	Own the decision

To use the check, go letter by letter and give each one a rating: green (no concerns), yellow (proceed with specific safeguards, written down) or red (don't proceed until it's fixed). Any red stops the project. Yellows become action items with owners and dates.

13.4 Sustainable marketing

The criticisms earlier in this chapter share a theme: practices that win a sale today at someone else's expense, whether that's the customer's, society's or the planet's. Sustainable marketing is the positive answer to that theme.

Sustainable marketing is broader than "green" marketing. It includes environmental impact, but also the long-term interests of customers and society. A company that sells products customers love but that harm their health, or that hits this quarter's target by burning out its dealer network, isn't marketing sustainably either.

13.4.1 Principles of sustainable marketing

Five principles guide sustainable marketing:

1. Consumer-oriented marketing. Organize marketing activities from the customer's point of view.
2. Customer-value marketing. Invest most resources in building real customer value rather than one-time promotions.
3. Innovative marketing. Keep looking for real product and marketing improvements.
4. Sense-of-mission marketing. Define the mission in broad social terms, not narrow product terms.
5. Societal marketing. Weigh consumers' wants, the company's requirements, consumers' long-run interests and society's long-run interests.

The last principle suggests a useful way to classify products by two questions: does it give customers immediate satisfaction, and does it serve their long-run interests?

	Low long-run benefit	High long-run benefit
High immediate satisfaction	Pleasing products: give pleasure now but may harm later (sugary drinks, some fast fashion)	Desirable products: satisfying now and beneficial later (a tasty, nutritious meal; a fun, efficient way to commute)
Low immediate satisfaction	Deficient products: neither appealing nor beneficial; should be fixed or dropped	Salutary products: good for people but less appealing (low-fat foods, seat belts, early versions of many green products)

The goal of sustainable marketing is to move products toward the desirable quadrant: make pleasing products healthier or less harmful, and make salutary products more appealing.

13.4.2 Sustainability in practice

Several well-documented companies show what sustainable marketing can look like:

• Patagonia ran a full-page advertisement on Black Friday 2011 with the headline "Don't Buy This Jacket," urging customers to think about the environmental cost of what they buy. Its Worn Wear program repairs and resells used Patagonia clothing, and in 2022 its founder transferred ownership of the company to a trust and a nonprofit dedicated to environmental causes. Patagonia's marketing works because its practices match its claims over many years.
• Repair, resale, take-back and redesign. Furniture, electronics and apparel companies now buy back, refurbish or recycle used products, and redesigns such as refillable packaging and repairable products cut waste at the source. Changing the product usually matters more than changing the message.

Sustainability also connects to the natural environment forces in Chapter 3. The Marketing Environment: resource costs, climate-related regulation and changing customer expectations make sustainability a strategic issue, not only an ethical one.

13.4.3 Greenwashing

As customers reward environmental claims, some companies make claims they can't back up.

The most famous case of environmental deception is Volkswagen's "clean diesel" scandal. In 2015 US regulators revealed that Volkswagen had installed software in diesel cars that detected emissions tests and reduced emissions only during testing. In normal driving, the cars emitted far more pollution than the law allowed, while the company marketed them as clean. The result was billions of dollars in fines and settlements, criminal charges and severe damage to the brand.

Most greenwashing is subtler. Common forms include:

Form	Example	What to do instead
Vague claims	"Eco-friendly," "green," "natural" with no specifics	Make specific claims you can prove ("frame made from 50% recycled aluminum")
Hidden trade-offs	Highlighting one green feature while ignoring bigger impacts	Describe the full picture, or at least don't imply it
No proof	Claims without evidence or certification	Substantiate claims before making them; cite independent certification
Irrelevant claims	Advertising the absence of a substance that's already banned	Claim only what's meaningful
Misleading imagery	Leaves, forests and green packaging on an ordinary product	Let the imagery match the substance
Offsets as erasure	"Carbon neutral" based mainly on purchased offsets	Explain what was reduced and what was offset

Regulators are paying attention. The FTC publishes guidance on environmental marketing claims (the "Green Guides"), and the European Union has adopted rules restricting vague environmental claims and requiring that green claims be substantiated. As with privacy, check current rules in each market.

13.5 Lumen Cycles: running the TRUST check

Dana decided to put each of Lumen's AI proposals through the TRUST check before her team spent any more time on them. She also asked for a sustainability claims review, since e-bikes are often marketed as a green alternative to driving. Here's what the team found.

13.5.1 Personalized offers from app ride data

The Lumen app tracks rides, battery health and service needs. Riders downloaded it for those features. The privacy notice in the app describes using ride data for service reminders and battery alerts, but it says nothing about marketing.

TRUST	Rating	Notes
Tell people	Yellow	Riders don't know ride data could drive offers. Must be explained clearly in the app.
Respect their data	Red	Using ride data for marketing is a new purpose riders never agreed to. Location patterns, such as daily trips to a school, are sensitive and can reveal where children are.
Unbiased outcomes	Yellow	Offers based on mileage could favor high-income suburban riders; test who receives which offers.
Straight dealing	Green	The offers themselves (accessories, upgrades, Lumen Care) are legitimate.
Take ownership	Yellow	Needs a named owner (Marcus) and a way for riders to see and change why they got an offer.

Decision: Not as proposed. The team redesigned it as an opt-in program: riders can choose "personalized tips and offers" in the app, with a plain explanation of what's used. Offers draw on mileage and battery data only, never on location patterns. The team dropped the idea of inferring "this rider has kids" from trip destinations altogether. As Dana put it, "If a rider wants Cargo information, they can tell us. We're not going to work it out from where they ride."

13.5.2 AI-generated lifestyle images

TRUST	Rating	Notes
Tell people	Yellow	Images are fine for clearly illustrative ads, but never in anything labeled as a real customer story.
Respect their data	Green	No customer data involved, as long as no real customer's photo is used as input.
Unbiased outcomes	Yellow	First drafts of generated "families" were strikingly similar: young, thin, mostly white, all in sunny weather. Needs review for range of family types, ages, bodies and abilities.
Straight dealing	Red	Some generated images showed the Cargo carrying three children and a full load of groceries, beyond the bike's rated capacity, and some showed details of the bike that don't exist.
Take ownership	Yellow	Jess owns image review; needs a checklist and a log of which images are AI-generated.

Decision: Approved with safeguards. Generated images may be used for social posts and display ads, labeled where the platform requires it, and every image is checked against the real product's specifications and load ratings. Testimonials and "meet a Lumen family" content use real customers, photographed with their permission.

13.5.3 Dynamic pricing on the website

TRUST	Rating	Notes
Tell people	Yellow	Customers don't expect a bike's price to change from visit to visit.
Respect their data	Yellow	Pricing on individual browsing behavior uses personal data for a purpose customers didn't expect.
Unbiased outcomes	Red	Pricing by location or device could charge more in areas with fewer dealers or competitors, which can line up with income and race.
Straight dealing	Red	Two neighbors paying different prices for the same bike on the same day would feel like a trick. Dealers, who sell at the suggested retail price, would also be undercut or embarrassed.
Take ownership	Yellow	No one on the team could yet explain how the proposed algorithm would set prices.

Decision: Rejected as proposed. The team kept the option of AI-assisted demand forecasting to time clear, public promotions (for example, an end-of-season sale on the Commute open to everyone), which passes the check. Prices for everyone who visits on a given day stay the same.

13.5.4 The website chat assistant

Lumen's existing chat assistant already hands warranty claims to a person. The check found two gaps: the assistant wasn't labeled as AI on every page, and its answers about battery range and load limits didn't always match the spec sheets. The fixes were a clear "AI assistant" label, a "talk to a person" button, safety answers drawn only from the official spec sheet, and a monthly review of sample conversations. After the Air Canada ruling, Dana didn't need convincing: anything the assistant tells a customer, Lumen has told the customer.

13.5.5 Sustainability claims

E-bikes have a real environmental story. A trip on an e-bike instead of a car usually means less fuel burned, less traffic and less parking. Lumen's mission, "Make the everyday trip cleaner, cheaper and more fun than driving," builds on that story. But Jess found several draft claims that wouldn't survive a greenwashing review:

• "Zero emissions" is misleading. Riding produces no tailpipe emissions, but manufacturing the bike and battery and generating the electricity to charge it do produce emissions. Better: "No tailpipe emissions," or a specific, sourced comparison with driving.
• "The greenest way to get around" is a vague superlative. Walking and conventional bikes have a better claim to it.
• "Planet-friendly batteries" has no proof behind it. Lumen's batteries have safety certification, which is a real, provable claim, but that's a safety claim, not an environmental one.

Batteries raised a bigger question: what happens to them at the end of their life? Lumen has no take-back program yet. The team proposed a battery take-back and recycling program run through its dealers and its Denver headquarters, so that customers could return old batteries safely when they replace them. The program would turn a weak claim into a real practice, and give customers one more reason to visit a dealer.

Your marketing plan. This week, draft the Responsible marketing review section of your plan. It should include:

1. An inventory of AI uses in your plan so far (in research, targeting, content, pricing, channels and customer service), and a TRUST check table for the two or three most important ones, with a rating and notes for each letter.
2. A data and privacy plan: what customer data the plan uses, where it comes from, what consent covers it, and what you will not collect or infer.
3. A claims review: every factual, safety and environmental claim in your communications and digital plans, the evidence behind each, and any you have changed or dropped.
4. A sustainability statement: one or two specific practices (not slogans) that make your offering more sustainable, and how you'll talk about them honestly.
5. A risk table listing the three biggest ethical or reputational risks in your plan, with the safeguard and owner for each.

You'll fold this section into the full plan in Chapter 14. Capstone: The Marketing Plan.

13.6 Try it with AI

In this exercise you'll use an AI assistant to create a small AI-assisted campaign, then audit it using the TRUST check. The goal is to practice spotting problems that look perfectly normal at first glance. Use only the fictional Lumen Cycles facts from this book, and don't paste any real customer or personal data into the tool.

Start by giving the AI the facts it needs and asking for a campaign.

Now audit the reply yourself before asking the AI anything else. Copy the campaign into a document and go through the TRUST check, letter by letter:

• Tell people. If the ads used the images described, would viewers assume they show real customers? Where would a disclosure be needed?
• Respect their data. What personal data would the targeting plan require? Is any of it sensitive, such as data that reveals someone has young children, their health or their precise location?
• Unbiased outcomes. Who is included and who is left out by the image descriptions and the targeting plan? Look for proxies: income levels, specific zip codes, age ranges, "affluent suburbs." Would single parents, grandparents raising children, families with a disabled member or families in apartments see themselves here?
• Straight dealing. Underline every factual claim. Which ones can you verify from the case facts? Mark any superlatives ("safest," "best"), environmental claims or performance claims the case doesn't support.
• Take ownership. If this campaign ran, what would a human need to review first, and who should be responsible?

Rate each letter green, yellow or red with a one-sentence reason. Then ask the AI to audit its own work.

Compare the AI's audit with yours. Make three lists:

1. Problems you both found.
2. Problems you found that the AI missed.
3. Problems the AI flagged that you missed, or that you think aren't real problems. For each, decide who is right and why.

AI tools are often better at critiquing text when asked than at avoiding problems in the first place. Notice whether that happened here, and whether the AI caught issues that require knowing the facts or only ones visible from wording alone.

Finally, ask for a revision.

Run the TRUST check on the revision. Did every red and yellow improve? Did the revision introduce anything new, such as a different unsupported claim? Write a short paragraph, about 150 words, on what this exercise taught you about using AI in campaign work, and keep your ratings tables. You can use them as a model for the TRUST check tables in your responsible marketing review.

13.7 Summary

Marketing faces real criticisms: that it raises prices, deceives and pressures buyers, sells unsafe or short-lived products, underserves disadvantaged consumers, fuels materialism and harms competition. Consumerism, environmentalism and regulation push back against these practices, but the most effective response is a company's own choice to build long-term customer value instead of short-term sales. Privacy is central to that choice. Responsible marketers seek specific, informed and reversible consent, collect only the data they need, avoid sensitive inferences and keep their data promises, because customer trust is easier to lose than to rebuild.

AI magnifies these issues. It can learn bias from historical data, persuade with manipulative precision, generate convincing fakes, hide its role from customers and blur who is responsible when it goes wrong. The TRUST check (Tell people, Respect their data, Unbiased outcomes, Straight dealing, Take ownership) turns those risks into five questions to ask before any AI-assisted practice goes live. Sustainable marketing extends the same long-term thinking to society and the environment: build products that are both satisfying and beneficial, and make only environmental claims you can prove, because greenwashing turns a strength into a liability.

13.8 Review Questions

1. List three criticisms of marketing's effects on individual consumers. For one of them, give the critics' argument and the marketers' response, and say which you find more convincing.
2. What makes consent "informed"? Explain the difference between opt-in and opt-out consent, and why the difference matters to the size and quality of a marketing audience.
3. A clothing retailer's AI ad system shows ads for its premium line mostly to people in high-income zip codes and ads for its discount line mostly to people elsewhere. Nobody told it to use income. Explain how this could happen and how you would test whether it's a problem.
4. Explain the difference between persuasion and manipulation. Give an example of an AI-enabled marketing tactic that crosses the line, and explain how you would change it so it doesn't.
5. Why did the Air Canada chatbot ruling matter for companies that use AI to talk with customers? Which letter of the TRUST check does it illustrate most directly?
6. A sunscreen brand plans to advertise its product as "reef-safe and eco-friendly" with images of a coral reef. Using the table of greenwashing forms, identify the risks and rewrite the claim so it's specific and supportable.
7. Apply the TRUST check to a use of AI you've encountered as a customer, such as a chatbot, a recommendation feed or a personalized price. Rate each letter and explain your ratings.
8. Lumen rejected dynamic pricing on its website but kept AI-assisted demand forecasting for public promotions. Do you agree with this decision? What would have to change for some form of personalized pricing to pass the TRUST check?