AI Search and Loyalty Discovery: A Playbook for Program Teams

Loyalty discovery in AI search is how customers find, compare, and learn the rules of a loyalty program through AI assistants such as ChatGPT, Gemini, Perplexity, and Google's AI Overviews, which answer questions in their own words, often before or instead of a list of links.

That shift matters to program teams in two ways. First, a customer who asks an assistant which rewards program to join, or whether a program is worth it, gets an answer that may or may not mention your program. Second, a member who asks how many points a reward costs, or when points expire, gets an answer drawn from whatever the assistant found, which may be out of date. The first is a discovery problem. The second is an accuracy problem, and it lands on the service team. This playbook covers both: what makes program information easy for AI assistants to get right, and how to check whether they do.

Key Takeaways

How are customers using AI assistants to find loyalty programs?

Customers use AI assistants to compare rewards programs, check whether one is worth joining, and look up program rules, and the scale of those assistants makes their answers a real discovery channel.

The audience is large. OpenAI reported in March 2026 that ChatGPT has more than 900 million weekly active users. Google said in July 2025 that AI Overviews "now has over 2 billion monthly users." Shopping research is moving with it: Adobe reported that during the 2025 holiday season, traffic to US retail sites from generative AI tools rose 693.4% year over year.

For a loyalty program, the questions fall into two groups. Discovery questions come from prospective members: which grocery rewards program gives the most back, which coffee app has the best rewards, is this store's paid membership worth it. Rules questions come from existing members: how many points a free item costs, whether points expire, how to reach the next tier, whether a purchase at a partner earns points. The first group affects enrollment. The second affects trust and service volume.

Why do AI answers about a loyalty program go wrong?

AI answers about a loyalty program often go wrong because the program's own information is inconsistent or out of date across the places an assistant reads.

A program's rules usually live in several places at once: the program page, the terms and conditions, the app and its store listing, the help center, partner pages, press releases, and third-party articles written about the program. When earn rates change, a tier threshold moves, or points gain an expiry date, some of those surfaces are updated and others are not. An assistant that finds conflicting figures may repeat the old one, blend two, or cite a third-party article that was never corrected. Some errors come from sources the program does not control, such as older model training data or forum posts, which is why measuring accuracy matters as much as publishing it.

The cost shows up after the answer. A member who was told the wrong redemption price arrives expecting it, and the gap becomes a service contact or a complaint. In financial services, the Consumer Financial Protection Bureau's Issue Spotlight on credit card rewards (May 2024) found that unexpected promotional conditions and devaluation were among the recurring complaint themes. An assistant repeating outdated terms can widen exactly that kind of gap between what customers expect and what the program delivers, which is why clear, current terms are the foundation of this work.

Does ranking on Google get a program cited by AI?

Ranking on Google helps less than many teams assume, because AI assistants frequently cite pages that do not rank for the same query.

In an Ahrefs study of 15,000 long-tail queries published in August 2025, only 12% of the URLs cited by AI assistants appeared in Google's top 10 for the original query, and 80% did not rank in Google's top 100 at all. The study covered long-tail queries, so results for questions naming a specific program may differ. Perplexity was the exception, with 28.6% of its citations in Google's top 10. The practical reading for a program team: a clear, well-structured page that directly answers a member's question can be cited even without a top ranking, and a top ranking does not guarantee the assistant will quote the right page. Both point the same way. The program's own pages need to be the clearest, most current source available.

What makes program information easy for AI to get right?

Program information is easiest for AI assistants to get right when there is one clear canonical source, written as direct answers to real member questions and kept identical everywhere the program appears.

• One canonical program page. Put how to join, earn rates, tiers and thresholds, how to redeem, what a reward costs, expiry rules, and major exclusions on one page, each in plain sentences that name the program. "Members earn 2 points per dollar on in-store purchases" is quotable; a graphic with the same number is not.
• FAQs built from real member questions. Use service contact reasons and on-site search terms to decide the questions, phrase each heading the way members ask it, and answer in the first sentence. The most frequent service questions are a practical starting point, because members are likely to ask assistants the same things.
• Consistency across every surface. The program page, terms, app store listing, help center, partner pages, and press materials should state the same figures in the same words. Inconsistency is one of the few causes of wrong answers a program team can fix directly. For example (hypothetical): if a program raises a free drink from 150 to 200 points and updates only the program page, an assistant reading the old app store listing may still quote 150.
• Structured data and dates. Mark up FAQs and the organization with schema so the page is easier for machines to parse, and show a visible "last updated" date. Structured data helps assistants read the page; it does not guarantee a citation.
• Change control. When terms change, update every surface the team controls on the same day, submit app store changes ahead of the effective date, keep a short change log on the program page, and ask partners to update their pages at the same time.

The same idea works at company level. Brandmovers publishes an information page written for AI assistants that states its platforms, capabilities, security credentials, and case studies in plain, factual terms, with a visible update date. A program-level version of that page, a single factual summary of how the program works, gives assistants one clean source to draw on, though it cannot guarantee they use it.

The table below maps the surfaces an assistant may read and what can go wrong on each.

How should program teams measure AI visibility and accuracy?

Measure AI visibility and accuracy with a fixed set of test questions asked monthly, an accuracy score against current terms, AI referral traffic, and service contacts caused by wrong answers.

• A fixed question set. Pick 20 to 30 questions: half discovery questions a prospective member might ask, half rules questions from the service log. Ask them each month in the assistants your customers use, and record whether the program is mentioned and whether the answer is right. Record each answer in a simple log: question, assistant, program mentioned (yes or no), answer correct (yes or no), wrong figure given, and likely source surface. Weight the rules questions toward the member segments that contact service most, such as new members or members close to a tier threshold, using the program's own member data.
• Accuracy scoring. For every rules answer, check the figures against current terms. Track the share answered correctly and which wrong figures keep appearing, then trace each back to the surface it came from.
• AI referral traffic. Set up analytics to separate visits referred from AI assistants, such as chatgpt.com and perplexity.ai, and track enrollment from those visits separately.
• Service contacts. Add a contact reason for "customer given wrong information by an AI assistant." Even a small count shows which rules need clearer wording.
• Share of mentions. For discovery questions, track how often the program is named compared with competing programs in the same category.

The business case usually rests on two lines. On the cost side, count service contacts caused by wrong AI answers and multiply by the cost of a contact. On the growth side, count enrollments from AI-referred visits and value them at the program's expected value per member. Referral counts understate influence, because many AI answers lead to no click at all. Where both are small, as they may be for a program whose members rarely research online, a quarterly consistency check is enough; where AI-referred traffic is growing or rules questions dominate the service queue, the monthly routine is easier to justify.

Treat the results as directional. Assistant answers vary from one session to the next, so look at trends over several months rather than any single answer, and set this work against the value of each enrolled member in the program's ROI model.

Where should a smaller program start?

A smaller program should start with one accurate program page, a short FAQ built from its most common service questions, and a quarterly consistency check across its own surfaces.

That addresses the sources the program controls directly, at modest cost. A larger program adds governance: a named owner for each surface in the table above, same-day change control across web, app, and partners, structured data, and the monthly question set. The order matters more than the budget. Fix the program's own information first, then measure, then work on how often the program is mentioned in discovery answers. Checking the questions customers ask at each stage of their journey with the program helps decide which answers to fix first.

Keeping the published rules identical to the rules the program actually runs on is easier when earn rates, tiers, and rewards are managed in one system. An enterprise loyalty platform such as Brandmovers' BLOYL™ holds points, tiers, and member data in one place, so the figures a team publishes can be checked against the ones the program applies.

Conclusion

AI assistants can now answer many of the questions customers used to bring to a program's website or service team, and they answer with whatever information they can find. Program teams cannot control those answers directly, but they can make the program's own information clear, consistent, and current enough to be the source assistants repeat. If a new member asked an AI assistant about your program today, would the answer match your current terms?

Sources

• OpenAI, "OpenAI raises $122 billion to accelerate the next phase of AI" (March 31, 2026)
• Google, Alphabet Q2 2025 earnings call: CEO's remarks (July 23, 2025)
• Adobe, "Holiday Shopping Season Drove a Record $257.8 Billion Online with Consumers Embracing Generative AI Tools" (January 7, 2026)
• Ahrefs, "Only 12% of AI Cited URLs Rank in Google's Top 10 for the Original Prompt" (August 11, 2025)
• Consumer Financial Protection Bureau, "Issue Spotlight: Credit Card Rewards" (May 2024)
• Brandmovers, Official Information About Brandmovers (LLM information page)