Ethics, Governance, AI, and the Future of Search

Introduction

Search marketing is no longer concerned only with rankings, keywords, advertisements, and conversion rates. It operates within a social, legal, technical, and economic environment in which organisations collect personal data, influence information access, automate decisions, and compete for users’ attention. These activities create responsibilities that extend beyond campaign performance.

Ethical search marketing seeks to create value without deception, unlawful surveillance, discrimination, manipulation, or avoidable harm. Governance converts these principles into policies, roles, controls, evidence, and review processes. Artificial intelligence (AI) makes the need for both more urgent: automated systems can improve relevance and productivity, but they can also reproduce bias, generate inaccurate content, expose confidential data, and make accountability difficult.

This chapter examines privacy and consent, accessibility, brand safety, click fraud, algorithmic bias, regulatory compliance, AI-generated search experiences, zero-click search, conversational discovery, and governance for sustainable search marketing.

Learning Objectives

By the end of this chapter, you should be able to:

1. Explain the ethical responsibilities associated with search engine optimisation (SEO) and search engine marketing (SEM).
2. Distinguish privacy, consent, transparency, and data-minimisation requirements.
3. Evaluate accessibility, brand safety, click fraud, and algorithmic bias risks.
4. Identify major regulatory and platform-governance considerations.
5. Assess the effects of AI-generated answers, zero-click search, and conversational discovery.
6. Design a practical governance framework for responsible search marketing.

1. From Search Performance to Responsible Influence

Search marketing influences what people see, trust, buy, and believe. A search result may affect healthcare choices, employment opportunities, financial decisions, political understanding, or access to public services. Consequently, a technically successful campaign may still be irresponsible if it relies on hidden surveillance, misleading claims, inaccessible design, or exploitative targeting.

A useful distinction is between performance legitimacy and social legitimacy. Performance legitimacy concerns whether a campaign achieves its stated objectives, such as qualified traffic or profitable sales. Social legitimacy concerns whether the campaign is lawful, fair, transparent, accessible, and consistent with stakeholder expectations. Sustainable search marketing requires both.

The main stakeholders are:

• Users, who provide attention, data, and sometimes money.
• Advertisers and publishers, who seek visibility and revenue.
• Search platforms, which organise information and allocate advertising access.
• Employees and agencies, who design and operate campaigns.
• Regulators and civil society, who protect competition, rights, safety, and public interests.
• Communities, which may experience environmental, cultural, or social effects from marketing practices.

Ethical analysis should therefore ask four questions:

1. Is the activity lawful?
2. Is it honest and understandable to the intended audience?
3. Could it create foreseeable harm or unfair disadvantage?
4. Can the organisation demonstrate how decisions were made and corrected when necessary?

2. Privacy, Consent, and Responsible Data Use

2.1 The search-data ecosystem

Search marketing can involve query terms, IP addresses, device identifiers, cookies, advertising IDs, location signals, browsing behaviour, purchase records, customer lists, and inferred interests. Some data may be directly supplied by a person; other data may be observed or inferred. A search for a sensitive medical condition, for example, may reveal information that a user did not intend to disclose to an advertiser.

Personal data generally means information relating to an identified or identifiable person, although the precise legal definition differs by jurisdiction. Pseudonymised data may still be regulated if a person can reasonably be re-identified. Aggregated data may have lower privacy risk, but aggregation is not automatically irreversible.

2.2 Core privacy principles

Responsible data practice commonly includes the following principles:

• Purpose limitation: collect and use data for specified, legitimate purposes.
• Data minimisation: collect only what is reasonably necessary.
• Accuracy: correct materially inaccurate records.
• Storage limitation: retain data only as long as justified.
• Security: protect data through access controls, encryption, monitoring, and incident response.
• Transparency: explain what is collected, why, with whom it is shared, and how choices can be exercised.
• Accountability: maintain evidence that the organisation has implemented these principles.

These principles appear in different forms in laws such as the European Union’s General Data Protection Regulation (GDPR), the UK GDPR and Data Protection Act 2018, the California Consumer Privacy Act as amended by the California Privacy Rights Act, and sector-specific laws in many other countries. Organisations must obtain current legal advice because applicability depends on location, sector, business role, data type, and changing legislation.

2.3 Consent is not the same as notice

A privacy notice informs a person. Consent is a legal or ethical permission mechanism with specific requirements. Meaningful consent should normally be informed, specific, freely given, and capable of withdrawal. A pre-ticked box, confusing interface, or bundled acceptance may not represent meaningful choice.

Consent is not always the only lawful basis for processing under data-protection regimes. Depending on the jurisdiction and activity, other bases may include contract, legal obligation, legitimate interests, or vital interests. However, a lawful basis does not remove duties of fairness, transparency, security, or minimisation.

Search marketers should distinguish:

2.4 Privacy by design in campaign operations

Privacy by design means considering privacy before a campaign is launched rather than treating it as a legal review at the end. A practical process is:

1. Map the data flow from collection to deletion.
2. Identify the purpose and minimum required fields.
3. Classify sensitive data and vulnerable audiences.
4. Select a lawful basis and document it.
5. Configure consent and opt-out controls.
6. Restrict access and vendor permissions.
7. Test retention, deletion, and user-rights procedures.
8. Review changes in platform policy and law.

A responsible measurement plan may favour aggregated conversion reporting, contextual targeting, first-party data collected transparently, and privacy-enhancing technologies rather than unrestricted individual tracking.

3. Accessibility and Inclusion

Search marketing is ethically defective if people cannot perceive, understand, navigate, or act on the content. Accessibility benefits users with disabilities and often improves usability for everyone, including mobile users, people with temporary impairments, older users, and people operating in noisy or low-bandwidth environments.

Important practices include:

• using meaningful page titles and headings;
• providing text alternatives for informative images;
• ensuring keyboard navigation;
• maintaining sufficient colour contrast;
• providing captions and transcripts for video;
• using descriptive link text;
• avoiding content that flashes dangerously or depends only on colour;
• supporting screen readers and responsive layouts;
• writing clear, plain-language content;
• ensuring forms and error messages are understandable.

The Web Content Accessibility Guidelines (WCAG), developed by the World Wide Web Consortium, provide widely used technical guidance. Legal obligations vary. Some organisations may also be covered by disability-discrimination, public-sector accessibility, consumer-protection, or procurement requirements.

Accessibility is relevant to SEO because search engines need interpretable content and users need usable landing pages. However, accessibility should not be justified only as a ranking tactic. People’s equal participation is the primary reason.

4. Truthfulness, Dark Patterns, and Brand Safety

4.1 Truthful representation

Search advertisements and organic content should represent products, prices, evidence, limitations, endorsements, and availability accurately. Common risks include:

• implying an official relationship that does not exist;
• using a competitor’s name deceptively;
• hiding material conditions in small or unclear text;
• presenting advertorial content as independent information;
• making unsupported health, environmental, financial, or performance claims;
• creating urgency that is false or artificially generated.

The principle of materiality is important: information is material when its omission or presentation could affect a reasonable person’s decision. Disclosures should be prominent, understandable, and placed where the claim is made.

4.2 Dark patterns

Dark patterns are interface designs that steer, pressure, confuse, or obstruct users in ways that favour the organisation over the user’s informed choice. Examples include difficult cancellation, disguised advertising, default consent, confirmshaming, hidden fees, and misleading buttons. A high conversion rate does not prove ethical success if conversion is achieved through manipulation.

4.3 Brand safety and suitability

Brand safety concerns the risk that advertising appears beside illegal, extremist, hateful, violent, sexually explicit, fraudulent, or otherwise harmful content. Brand suitability is broader: it considers whether placement fits the organisation’s values, audience, sector, and risk tolerance.

Controls may include keyword exclusions, topic exclusions, placement lists, inventory-quality standards, verification vendors, contextual review, frequency controls, and escalation procedures. No automated classification system is perfect. Human review is especially important for breaking news, satire, minority-language content, and rapidly changing events.

A governance policy should define unacceptable content, tolerable risk, approval authority, monitoring frequency, and the process for pausing campaigns. It should also consider the opposite risk: excessive blocking can exclude legitimate journalism, educational content, or marginalised perspectives.

5. Click Fraud and Invalid Traffic

5.1 Meaning and sources

Click fraud is intentional manipulation of advertising interactions to generate improper benefit or cost. Invalid traffic is a broader platform term that may include automated traffic, accidental clicks, incentivised activity, bots, malware, click farms, competitor activity, or technically abnormal behaviour. Not all invalid traffic is malicious, and not every suspicious click can be identified with certainty.

Potential consequences include wasted budget, distorted performance data, unfair competition, and incorrect optimisation decisions. Search marketers should not accuse a publisher or competitor solely from unusual traffic patterns.

5.2 Detection and response

Useful signals may include abnormal click-through rates, repeated interactions from concentrated sources, implausible time patterns, high click volume with no meaningful engagement, inconsistent geographic or device data, and conversion rates that differ sharply from verified business outcomes. These are indicators, not proof.

A proportion metric can be used descriptively:

The numerator is the number of interactions classified as invalid under a documented rule, and the denominator is the total measured interactions for the same period and scope. The measure is meaningful only when detection methods, time period, attribution rules, and data sources are consistent. It is not a universal estimate of all fraud.

Illustrative example: If 1,200 of 40,000 measured clicks are classified as invalid, the rate is 1,200 ÷ 40,000 × 100% = 3%. The organisation should then examine financial impact, evidence quality, platform credits, traffic sources, and whether valid users were incorrectly excluded.

Good practice includes using platform invalid-traffic reporting, server-side logs where lawful, conversion-quality checks, bot-management tools, anomaly alerts, independent verification, and documented dispute procedures. Blocking entire countries, devices, or demographic groups without evidence can create unfair exclusion.

6. Algorithmic Bias and Search Fairness

Search ranking, ad delivery, audience selection, content moderation, and automated bidding all involve algorithms. Bias can enter through training data, labels, proxies, business objectives, historical inequalities, feedback loops, or unequal access to technology.

Examples include:

• lower visibility for dialects or minority-language content;
• stereotypical associations between occupations and demographic groups;
• delivery of high-value adverts unevenly across groups;
• automated moderation that misclassifies legitimate reclaimed language;
• ranking systems that favour already dominant publishers;
• facial, location, or behavioural inferences that disadvantage vulnerable people.

Bias is not limited to intentional discrimination. A system can produce disparate outcomes while using apparently neutral variables. At the same time, unequal outcomes do not automatically prove unlawful discrimination; context, justification, data quality, and legal tests matter.

A practical fairness assessment should ask:

1. Who benefits and who may be excluded?
2. Which variables act as direct or indirect proxies for protected characteristics?
3. Are training and evaluation data representative?
4. Do error rates differ among relevant groups?
5. Can affected people obtain an explanation or challenge a decision?
6. Is there human oversight with authority to intervene?

Where sufficient lawful data exist, teams may compare outcome rates across groups. A simple selection-rate ratio is sometimes called a disparate-impact or four-fifths-style indicator:

This is a screening measure, not a universal legal test. It requires carefully defined groups, comparable opportunities, valid denominators, adequate sample sizes, and statistical and legal interpretation. Organisations should avoid collecting sensitive attributes merely to calculate a metric without a lawful and ethical basis.

Mitigation can include better representative data, removal of unjustified proxies, constraint-based optimisation, subgroup testing, human review, accessible appeals, supplier audits, and continuous monitoring. Fairness may involve trade-offs with accuracy, privacy, and commercial efficiency; those trade-offs should be explicitly governed rather than hidden.

7. Regulatory and Platform Compliance

Search marketing may be affected by several overlapping regulatory areas:

• data protection and privacy;
• consumer protection and unfair commercial practices;
• advertising disclosure and endorsement rules;
• competition and antitrust law;
• intellectual property and copyright;
• accessibility and equality law;
• children’s and vulnerable-person protections;
• cybersecurity and electronic communications;
• sector-specific rules for health, finance, gambling, employment, and political advertising.

The relevant law depends on the organisation’s location, the audience’s location, the service, and the parties involved. A platform’s policy is not a substitute for law, and legal compliance is not a complete substitute for ethical responsibility.

A compliance register should identify each requirement, owner, evidence, review date, affected campaign, and escalation path. Evidence may include consent records, creative approvals, claim substantiation, accessibility tests, vendor contracts, risk assessments, incident records, and deletion logs.

The three lines model is useful for governance:

1. First line: marketing and operational teams own daily risks and controls.
2. Second line: privacy, legal, compliance, security, or risk specialists advise and monitor.
3. Third line: internal audit or independent assurance evaluates the system.

Smaller organisations may combine roles, but they should not eliminate accountability. Someone must have authority to stop a campaign.

8. AI in Search Marketing

8.1 Uses of AI

AI can support keyword clustering, content briefs, translation, bid optimisation, anomaly detection, audience modelling, customer-service responses, creative variation, structured-data generation, and search-result summarisation. These uses can reduce repetitive work and improve analysis.

However, AI outputs may be inaccurate, biased, non-original, insecure, or inconsistent with brand and legal requirements. Generative systems can invent sources, create plausible but false claims, reproduce copyrighted material, expose confidential prompts, or generate content at a scale that reduces information quality.

8.2 Human accountability

The organisation remains responsible for published claims and campaign outcomes even when a platform or vendor supplied the model. Human review should be proportionate to risk. A low-risk title variation may require sampling; a medical claim, financial recommendation, political message, or legal statement requires specialist verification before publication.

An AI-use policy should specify:

• approved and prohibited use cases;
• information that must not be entered into public tools;
• ownership and review responsibilities;
• source verification requirements;
• disclosure expectations;
• intellectual-property and licensing checks;
• testing for bias, hallucination, and security weaknesses;
• logging, retention, and incident response;
• conditions for human approval and withdrawal.

8.3 AI-generated content and search quality

Automation does not inherently make content unacceptable. The key questions are whether content is accurate, useful, original enough for its purpose, transparently produced where disclosure matters, and edited by someone accountable. Producing large volumes of low-value pages solely to manipulate rankings is inconsistent with sustainable SEO and may violate platform spam policies.

9. AI-Generated Search Experiences and Zero-Click Search

Traditional search commonly presented a list of links. Newer experiences may produce summaries, answer boxes, conversational responses, multimodal results, or direct actions. Users may obtain an answer without visiting a website; this is known as zero-click search.

Zero-click search changes the meaning of visibility. A brand may gain recognition or influence without receiving a click, while publishers may lose traffic despite being used as sources. Search marketers should therefore measure more than sessions. Relevant indicators may include:

• qualified visits and conversions;
• branded-search demand;
• assisted conversions;
• citation or source inclusion where observable;
• direct traffic and returning users;
• lead quality and customer outcomes;
• sentiment, trust, and error reports;
• accessibility and complaint rates.

No single metric captures visibility in AI-generated answers. Platform reporting may be incomplete, non-comparable, or subject to change. Organisations should avoid presenting estimated answer visibility as precise fact.

To improve the probability that information is understood accurately, organisations should maintain clear, authoritative, crawlable content; consistent facts across trusted sources; structured data where appropriate; strong editorial standards; identifiable authorship; and accessible pages. These practices do not guarantee inclusion or control how a model summarises content.

There are also important rights and governance questions: how content is collected for training or retrieval, whether creators receive attribution, whether summaries are accurate, how corrections are made, and who is responsible when an answer causes harm. Search marketers should treat citations, licensing, robots directives, publisher preferences, and platform terms as part of an evolving governance environment.

10. Conversational Discovery and the Changing Customer Journey

Conversational discovery allows users to express complex goals rather than enter short keyword strings. A user might ask for a comparison, explain constraints, request a plan, or refine an answer through dialogue. This creates opportunities for brands to provide expertise, tools, and customer support, but also risks over-personalisation, fabricated confidence, and dependency on an intermediary.

The customer journey may become less linear:

1. A user expresses a need conversationally.
2. An AI system interprets intent and constraints.
3. It retrieves, ranks, or generates options.
4. The user asks follow-up questions.
5. A transaction or referral occurs through the interface.
6. The user evaluates the result after use.

Marketing implications include investing in genuine expertise, structured product information, high-quality first-party content, accessible service interfaces, and trust signals. Organisations should not design conversational agents to conceal commercial relationships or pressure users into decisions. Users should know when they are interacting with an automated system and how to reach a human or independent information source when appropriate.

11. A Governance Framework for Sustainable Search Marketing

Governance is the system through which an organisation directs, controls, documents, and improves search activity. It should be practical rather than a collection of abstract principles.

11.1 Governance components

A mature framework includes:

• Principles: privacy, truthfulness, inclusion, safety, accountability, and sustainability.
• Risk classification: low, medium, and high-risk activities based on audience, data, claims, automation, and potential harm.
• Roles: accountable executive, campaign owner, privacy or compliance adviser, technical owner, reviewer, and incident lead.
• Policies: data use, accessibility, AI, advertising claims, brand safety, vendors, and crisis response.
• Controls: approval gates, permissions, exclusions, testing, monitoring, and audit logs.
• Metrics: performance, quality, fairness, privacy, accessibility, complaints, and incident measures.
• Assurance: periodic review, independent testing, supplier assessment, and corrective action.
• Remedy: correction, takedown, refund, opt-out, appeal, notification, or other appropriate response.

11.2 Risk-based approval

Not every activity requires the same level of scrutiny. A campaign promoting a general clothing product to a broad audience may be lower risk than a campaign using sensitive health inferences to promote financial or medical services. Risk increases with sensitivity, scale, automation, opacity, vulnerability, irreversibility, and potential impact.

A simple qualitative risk matrix can classify likelihood and impact as low, medium, or high. It should not be mistaken for a mathematically precise probability model. High-risk activity should require documented justification, specialist review, testing, monitoring, and a clear stop mechanism.

11.3 Sustainable operating cycle

A responsible search programme can follow this cycle:

Plan → assess risk → obtain approval → implement → monitor → investigate → remediate → learn.

The cycle should include pre-launch testing and post-launch review. A campaign that performs well but generates complaints, accessibility failures, or discriminatory delivery should be corrected rather than celebrated.

12. Illustrative Case: Health-Information Search Campaign

Background

A fictional private clinic, WellPath, wants to attract patients searching for information about a chronic condition. It proposes search ads, remarketing, AI-generated educational pages, and an automated chatbot.

Problem

The proposed campaign could reveal sensitive interests, make unsupported health claims, target vulnerable people aggressively, and provide inaccurate AI-generated answers. The clinic also wants to maximise conversions through a short online form.

Analysis

The campaign has high privacy and safety risk because health-related searches may be sensitive and users may be vulnerable. The organisation must distinguish general education from diagnosis or treatment advice. It must also consider whether remarketing could expose a person’s interests to others sharing a device. The chatbot could create harm if it gives confident but incorrect advice or fails to direct urgent cases to professional help.

Approach and solution

WellPath adopts the following controls:

• It conducts a documented privacy and risk assessment.
• It uses contextual search advertising rather than sensitive individual profiles where feasible.
• It provides clear privacy information and a genuine choice about optional tracking.
• It minimises form fields and separates appointment information from unnecessary profiling.
• Every health claim is reviewed by a qualified professional and supported by reliable sources.
• AI-generated drafts are never published without human fact-checking and editorial approval.
• The chatbot identifies itself as automated, avoids diagnosis, recognises urgent-risk language, and provides a human or emergency referral pathway.
• Landing pages meet accessibility requirements and are tested with assistive technologies.
• Campaign delivery, complaints, errors, and conversion quality are monitored.
• The clinic establishes a rapid pause and correction procedure.

Outcome

The campaign produces fewer raw leads than an unrestricted targeting strategy but a higher proportion of appropriate appointments. The clinic reduces privacy exposure, improves trust, and gains a documented basis for responding to regulators and patients. The case demonstrates that responsible governance may change the definition of success from maximum volume to sustainable value.

Learning points

1. Sensitive contexts require stronger controls than ordinary product promotion.
2. Human accountability cannot be delegated to an AI tool.
3. Privacy, accessibility, accuracy, and conversion quality can reinforce one another.
4. A documented decision process is itself an organisational asset.

13. Challenges and Limitations

Ethical governance has practical limits. Laws differ across countries and may be ambiguous. Platform systems are often opaque, making independent verification difficult. Small businesses may lack specialist staff. Privacy-preserving measurement can reduce detail. Bias testing may be constrained by limited or legally sensitive demographic data. Brand-safety filters can over-block legitimate content. AI systems change faster than internal policies and procurement cycles.

There are also genuine value conflicts. Personalisation may improve relevance but increase surveillance. Fraud controls may reduce abuse but block legitimate users. Content moderation may protect safety but suppress expression. Automated optimisation may improve efficiency but reduce explainability. These conflicts cannot always be solved by one universal rule. They require context, proportionality, stakeholder participation, evidence, and periodic review.

14. Best-Practice Checklist

Before launch, a search team should be able to answer “yes” to most of the following questions:

• Is the purpose of data collection specific and documented?
• Are consent, opt-out, retention, and deletion processes functional?
• Are claims accurate, evidenced, and appropriately disclosed?
• Can users with disabilities access and complete the journey?
• Have brand-safety exclusions and escalation procedures been tested?
• Are invalid-traffic indicators monitored without unsupported accusations?
• Has possible algorithmic bias been considered and tested?
• Has the AI tool been approved for the information being processed?
• Is there human review for high-risk content and decisions?
• Can the organisation explain, correct, pause, and audit the campaign?
• Are vendors contractually and operationally accountable?
• Does success include trust, quality, inclusion, and long-term value rather than clicks alone?

Chapter Summary

Ethical search marketing combines lawful conduct, honesty, privacy protection, accessibility, fairness, safety, and accountability. Privacy requires purpose limitation, minimisation, transparency, appropriate lawful grounds, security, and meaningful user choice. Accessibility ensures that search journeys can be used by people with diverse abilities. Brand safety, truthful claims, and avoidance of dark patterns protect users and organisational trust. Click fraud and invalid traffic require evidence-based detection rather than unsupported accusations.

Algorithmic bias can arise from data, objectives, proxies, feedback loops, and unequal system effects. It should be assessed through representative testing, human oversight, explanation, appeal, and corrective action. AI can improve search operations but introduces risks of hallucination, bias, privacy leakage, insecure use, and unaccountable automation. AI-generated answers, zero-click search, and conversational discovery require broader measures of visibility and stronger attention to source quality, attribution, accuracy, and user agency.

Governance turns principles into roles, risk classifications, policies, controls, monitoring, assurance, and remedy. Sustainable search marketing is therefore not simply the pursuit of maximum traffic. It is the creation of durable value while respecting rights, reducing foreseeable harm, and remaining able to explain and correct organisational decisions.

Key Terms

Accessibility: The design of content and systems so that people with diverse abilities can perceive, understand, navigate, and use them.

Algorithmic bias: Systematic and potentially unfair differences in algorithmic outcomes associated with data, design, or deployment.

Brand safety: Measures that prevent advertising from appearing beside content likely to cause serious reputational or social harm.

Brand suitability: The broader assessment of whether an advertising environment fits an organisation’s values and risk tolerance.

Click fraud: Deliberate manipulation of advertising interactions for improper benefit.

Consent: A person’s informed, specific, freely given, and withdrawable permission where consent is the applicable basis.

Conversational discovery: Finding information, products, or services through interactive natural-language dialogue.

Data minimisation: Limiting collection and use to what is necessary for a specified purpose.

Dark pattern: A design that manipulates or obstructs informed user choice.

Governance: The structures and processes used to direct, control, document, and improve organisational activity.

Invalid traffic: Traffic or interactions considered illegitimate, abnormal, automated, or otherwise unsuitable for reliable advertising measurement.

Lawful basis: The legally recognised justification for processing personal data.

零-click search: A search experience in which the user obtains an answer or action without visiting a result page.

Privacy by design: Integrating privacy considerations into systems and processes from the beginning.

Remarketing: Advertising to users based on prior interaction with a site, service, or brand.

Selection-rate ratio: The selection rate for one group divided by the selection rate for a reference group.

Transparency: Clear communication about data practices, commercial relationships, automated systems, and decision processes.

Zero-click search: A search experience in which the user obtains an answer or action without visiting a result page.

Review Questions

Short Answer Questions

1. What is the difference between a privacy notice and consent?
2. Define brand safety and explain how it differs from brand suitability.
3. Identify three principles of responsible data use.
4. Why is accessibility relevant to search marketing?
5. What is zero-click search?

Long Answer Questions

1. Explain how privacy by design can be applied to a remarketing campaign.
2. Discuss the main causes and mitigation strategies for algorithmic bias in search advertising.
3. Evaluate the benefits and risks of using generative AI to create SEO content.
4. Explain how governance converts ethical principles into operational controls.
5. Discuss how AI-generated search answers may change the measurement of SEO success.

Analytical / Application Questions

1. A retailer observes a sudden increase in clicks but no corresponding increase in qualified conversions. Design an evidence-based investigation into possible invalid traffic, including limitations of the evidence.
2. A financial-services advertiser wants to use behavioural data to target people who appear worried about debt. Identify the privacy, fairness, transparency, and consumer-protection risks, and propose a safer approach.
3. An organisation’s chatbot gives an inaccurate answer about a health service and the answer is reproduced in an AI search summary. Create an incident-response and governance plan covering correction, user protection, accountability, and future prevention.

Multiple Choice Questions with Answers

Question 1

Which practice best reflects data minimisation?

A. Collecting every available user attribute for possible future use

B. Collecting only the data necessary for a defined campaign purpose

C. Retaining all tracking data indefinitely

D. Sharing raw customer data with every advertising vendor

Correct Answer: B. Collecting only the data necessary for a defined campaign purpose

Question 2

Which statement best distinguishes consent from a privacy notice?

A. A privacy notice always grants permission to process data

B. Consent is only a technical cookie setting

C. A notice provides information, while consent is a permission mechanism where required

D. Consent removes the need for security controls

Correct Answer: C. A notice provides information, while consent is a permission mechanism where required

Question 3

What is the primary purpose of brand-safety controls?

A. To guarantee the lowest advertising cost

B. To prevent advertising from appearing beside harmful or unsuitable content

C. To increase the number of keywords in an account

D. To eliminate all organic search competitors

Correct Answer: B. To prevent advertising from appearing beside harmful or unsuitable content

Question 4

If 1,200 of 40,000 clicks are classified as invalid, what is the invalid-traffic rate?

A. 0.3%

B. 3%

C. 30%

D. 33.3%

Correct Answer: B. 3%

Question 5

Which is a limitation of using a selection-rate ratio to screen for possible bias?

A. It can never compare two groups

B. It automatically establishes a legal violation

C. It requires valid group definitions, denominators, and contextual interpretation

D. It measures website loading speed

Correct Answer: C. It requires valid group definitions, denominators, and contextual interpretation

Question 6

Which statement about AI-generated marketing content is most accurate?

A. The organisation is not responsible for content generated by a vendor

B. AI output should be published without review if it sounds plausible

C. AI can assist production, but high-risk claims require accountable human review

D. AI-generated content is always prohibited in SEO

Correct Answer: C. AI can assist production, but high-risk claims require accountable human review

Question 7

What does zero-click search mean?

A. A search with no ranking algorithm

B. A search in which the user obtains an answer or action without visiting a result page

C. A campaign that receives no impressions

D. A page that contains no links

Correct Answer: B. A search in which the user obtains an answer or action without visiting a result page

Question 8

Which governance role normally owns daily campaign controls?

A. The first line, consisting of operational and marketing teams

B. Only an external regulator

C. The third line, consisting exclusively of customers

D. An automated bidding system with no human owner

Correct Answer: A. The first line, consisting of operational and marketing teams

Question 9

Which design is most consistent with accessibility?

A. Using colour alone to communicate errors

B. Providing captions for important video content

C. Preventing keyboard navigation

D. Using vague link text such as “click here” for every link

Correct Answer: B. Providing captions for important video content

Question 10

Why should unusual click patterns not automatically be treated as proof of fraud?

A. All clicks are necessarily valid

B. Indicators can have innocent explanations and require evidence-based investigation

C. Fraud cannot affect advertising campaigns

D. Platforms never provide traffic data

Correct Answer: B. Indicators can have innocent explanations and require evidence-based investigation