Marketers have learned to count brand mentions, reviews, influencer reach, promo-code conversions, and comment sentiment. All of this creates the impression that word of mouth, the recommendations and conversations consumers have about a brand, has finally become measurable.
The problem is that digital dashboards capture only part of reality. A person may choose an insurer after speaking to a colleague, a doctor on a relative’s advice, or a contractor because a neighbor recommended one. These conversations may determine the purchase while leaving no social media mention, website review, or click for an attribution system to record.
This article is neither a rewrite nor a line-by-line summary of an academic paper. It is a practical synthesis of its main ideas, supplemented with our comments for marketers and business leaders. It is based on the article by professors Koen Pauwels and Zeynep Aksehirli, “Free and paid word-of-mouth from physical to digital, AI and beyond,” published on April 25, 2026, in the Journal of the Academy of Marketing Science. The complete paper is open access: read it on Springer or open the DOI page.
The authors synthesize research published since a foundational 2009 study of online word of mouth and propose a simple but useful matrix. It divides WOM along two dimensions: whether the person receives an incentive for the recommendation, and whether the communication happens online or offline.
Online conversations are attractive to analysts. Social listening systems can collect them, classify them by sentiment, subject, platform, and author, and compare them with advertising activity and sales. A conversation over coffee has no URL, timestamp, or UTM tag. That makes it easy to commit a methodological error: treating what can be measured as the entire universe of recommendations.
Pauwels and Aksehirli refer to research covering 500 brands in 15 industries. Online and offline WOM were both associated with sales, but the two correlated strongly mainly for high-tech brands. In most other categories, the relationship was close to zero. In other words, a loud online discussion does not guarantee that people discuss a brand just as often in person, and the reverse is also true.
The difference is not only about channels. People talk online and offline for different reasons. A social media post may demonstrate taste, knowledge, or membership in a particular group. In a private conversation, care, trust, reciprocity, and the desire to give advice suited to another person’s circumstances may matter more.
Fashion, travel, technology, alcohol, fitness, and premium products therefore tend to generate more public online WOM: they are easy to share as part of a person’s identity. Insurance, healthcare, home improvement, and household services may be less photogenic, yet they often require a private, contextual recommendation: whom to choose, whom to avoid, what the real cost is, and whether the provider can be trusted.
This is not a universal law for every brand. The authors themselves show that findings across studies sometimes conflict. A category provides a starting hypothesis, but the real balance of channels must be tested in the relevant market.
The authors’ matrix contains four cells:
| Organic recommendations | Paid or incentivized recommendations |
|---|---|
| Online | |
| Reviews, organic mentions, customer posts, and community discussions | Influencers, sponsored posts, affiliate marketing, and paid creator content |
| Offline | |
| Personal advice and spontaneous recommendations from friends, colleagues, and professionals | Referral rewards, ambassador programs, and incentivized personal recommendations |
The word “paid” should be interpreted broadly. The incentive may be money, a free product, a discount, a bonus, additional cloud storage, or some other benefit. Traditional definitions of WOM emphasized its non-commercial nature, so the term paid WOM may appear contradictory. The authors deliberately broaden the framework so that organic conversations can be compared with recommendations initiated or stimulated by a brand.
This group includes reviews, unpaid brand mentions, customer posts, videos describing product use, and discussions in specialist communities. Such content may remain discoverable for years and influence people well beyond the author’s immediate social circle.
Its main advantages are potential authenticity and scale. Yet organic does not mean positive or representative. People with exceptionally good or bad experiences are often the most likely to post. Some contributors may have hidden motives, and every platform exposes only a particular segment of the audience.
A familiar European example is IKEA. Distinctive room displays, storage solutions, and customer interiors naturally generate photographs and posts on Instagram, Pinterest, and other visual platforms. IKEA has formal terms allowing it to request permission to reuse selected user-generated images. In this case, the product and the experience give customers something worth showing. Still, the number of posts cannot tell IKEA how many purchases resulted from private advice or whether long-term loyalty changed.
Amazon provides another useful example. Its product pages turn dispersed customer experiences into a visible information layer built from ratings, written reviews, images, video, and Verified Purchase labels. For a shopper who cannot inspect the product physically, this type of WOM partly replaces asking a knowledgeable friend. But it is not a neutral census of all buyers: only a subset leaves reviews, Amazon’s rules shape what remains visible, and both verified and unverified perspectives can appear. Brands should therefore examine recurring themes and use cases, not just the average star rating.
Organic offline WOM includes face-to-face conversations, phone calls, and personal recommendations given without direct compensation. Its scale is difficult to measure, but the message is often tailored to its recipient. A friend does not merely name a medical practice; they explain which doctor they saw, what happened during the appointment, and who might not find that practice suitable.
This channel is especially important for small and midsize service businesses in the United States and Europe: dental practices, contractors, tutors, legal advisers, accountants, and local B2B providers. A dental practice may see strong Google Ads performance and few social mentions, while many new patients actually arrive after hearing, “I went there and they treated me well.” If the receptionist records those patients only as direct or phone leads, the marketing report systematically understates referrals.
Offline WOM is not magically more truthful. A person may be mistaken, repeat an outdated experience, or recommend an acquaintance for personal reasons. Its strength comes not from guaranteed accuracy but from the social relationship and the ability to provide contextual advice.
Paid online WOM includes influencer partnerships, sponsored posts, affiliate marketing, and creator content produced in return for money, products, or other benefits. A brand can quickly obtain reach, select an audience, establish a schedule, and measure at least some of the response.
The cost of control is a risk to authenticity. If a recommendation sounds like an advertising script, conflicts with the creator’s usual content, or conceals a commercial relationship, the audience may find it less convincing. This does not mean influencer marketing is ineffective. The paper discusses studies in which paid WOM increased sales and stimulated subsequent organic conversations. Outcomes depend on creator fit, category, form of compensation, disclosure, and genuine experience with the product.
Paid online WOM has a natural advantage in trend-driven categories because the product is easy to demonstrate, discuss, and purchase through a link. Yet large reach does not establish whether trust changed or whether attributed sales were incremental rather than simply assigned to the last promo code.
Paid offline WOM is a personal recommendation in which the sender or recipient receives a benefit. It includes referral programs, local ambassadors, incentivized recommendations, and some forms of in-person promotion.
Refer-a-friend programs used by European fintech companies, telecom operators, subscription services, and US digital products make the distinction easier to see. The same referral may belong to different cells of the matrix. When someone posts a referral link publicly, it is paid online WOM. When they explain a service to a colleague over lunch and then help the colleague register through the link, it is paid offline WOM with a digital conclusion.
Revolut is a clear European example. Eligible customers can receive a campaign invitation, send a unique referral link to someone they know, and earn a reward when the invitee completes specified steps. The company can record the link and completed actions, but it cannot fully observe the conversation that preceded them. One customer may forward the link without context; another may explain the product, demonstrate the app, and help a friend sign up. These referrals may look identical in a report even though the recommendation and transfer of trust were very different.
The lesson is that classification should reflect the communication between people, not only the technical conversion path. The existence of a referral link does not make the entire recommendation online.
In simplified terms, paid WOM buys predictability and scale, while organic WOM benefits from greater perceived authenticity. It would nevertheless be wrong to label one bad and the other good.
Organic recommendations cannot be launched according to a media calendar. A brand can create a product and experience worth discussing, but it cannot order customers to recommend it tomorrow. Paid WOM can quickly reach a relevant audience, support a launch, or explain a new product. Sometimes paid communication creates the first experience that later produces organic discussion.
The opposite risk is equally real. Too many identical integrations turn a recommendation into an ordinary ad format. A reward strictly tied to a particular behavior may undermine perceived sincerity more than free product access without a requirement to repeat a prepared script.
The right question is therefore not, “Which is better, organic or paid?” It is, “What role should each type of WOM play in this category, and how do they affect one another?”
Volume is easy to count, but one thousand mentions can represent very different situations: a wave of recommendations, a scandal, discussion of an ad, a mass question about what happened, or neutral statements about a purchase.
The authors distinguish at least four WOM dimensions: volume, valence, dispersion across people, and conversation topics. A management decision requires more than knowing how much people talk. It requires understanding who is speaking, to whom, about what, where, and with what consequence.
Even positive sentiment is not always the central factor. A neutral statement such as “I ordered mine there” may encourage observational learning: someone sees another person’s behavior and interprets it as evidence of normality or popularity. Negative discussion can occasionally raise awareness of a little-known brand, although deliberately manufacturing controversy would be a dangerous general strategy.
For a CMO, the practical conclusion is straightforward: share of mentions cannot be placed alongside sales without examining sources, content, and causes. Comparisons become especially misleading when one brand operates in a public lifestyle category and another sells a complex service through private consultations.
No single perfect counter exists. Measurement requires several sources, each illuminating a different part of the system.
Organic online WOM can be assessed through social listening, reviews, unpaid mentions, UGC, search queries, topics, sentiment, and the share of independent contributors. A measurement system should remove the brand’s own posts, paid content, automated reposts, and duplicates when the objective is to estimate earned conversation.
Paid online WOM analysis should go beyond reach, views, and engagement to include audience quality, visits, promo-code use, post-view and post-click behavior, brand lift, and control-group results. An advertising dashboard or influencer platform reports attribution inside its own ecosystem; it does not by itself prove incremental impact.
Organic offline WOM must be measured indirectly. Useful tools include post-purchase surveys, the questions “How did you first hear about us?” and “Who influenced your choice?”, verbatim responses, customer diaries, brand tracking, interviews, call-center data, and structured CRM fields. NPS alone is insufficient because willingness to recommend is not the same as an actual recommendation.
Paid offline WOM can be tracked through referral IDs, codes, CRM labels, geographic tests, control groups, and customer acquisition cost that includes rewards for both parties. Marketers should also monitor fraud, the cannibalization of organic recommendations, and cases in which a bonus is paid for a customer who would have converted anyway.
The best dashboard does not necessarily compress all four types into one figure. Its job is to show their relative roles, movement over time, measurement gaps, and connection to business outcomes.
The most interesting part of the paper concerns emerging questions rather than settled answers. The authors explicitly note that evidence about generative AI and spatial WOM remains limited, so their projections are speculative.
The first challenge is AI as the author of a recommendation. Generative systems can already produce reviews, comments, and influencer scripts at scale. This reduces the cost of creating WOM-like content while making its origin harder to determine: does the text describe an actual experience, or is it synthetic advertising?
The second challenge is AI as an intermediary. People increasingly ask an AI assistant rather than a friend or search engine which laptop to buy, which medical practice to visit, or which service is reliable. The model constructs its answer from a digital corpus in which online WOM is far better represented than private offline conversations. If a category traditionally depends on personal recommendations, the digital picture may be systematically incomplete.
The third challenge is the virtual influencer. Here, the very concept of a sender becomes blurred: the audience interacts with a character that has no conventional personal product experience. It is too early to declare such recommendations more or less effective. Brands already need, however, to consider disclosure, authorship, trust, and the way AI systems may distinguish an organic recommendation from a paid one.
Start not by launching another influencer campaign but by auditing the four cells. Where does the brand already receive recommendations? Which conversations are stimulated by money or benefits? What happens in person and never reaches digital reports? Which customers and experts actually influence decisions?
Next, connect each type of WOM to a business function. Paid online activity can create rapid reach during a launch. Organic online WOM can accumulate searchable evidence of experience. Organic offline WOM can reduce perceived risk in a complex or sensitive purchase. Paid offline WOM can activate an existing customer base or local community. These are hypotheses to test, not permanent assignments.
Then identify imbalance. If all analysis is based on social platforms, the brand may underestimate offline influence. If a business survives on referrals but records nothing, it does not know which customers, situations, and product attributes trigger those conversations. If a campaign relies exclusively on paid WOM, stopping the budget may immediately remove all visible activity.
Finally, assess transitions between the cells, not only the output of each one. Does influencer content generate organic discussion? Does a positive experience become a real recommendation? Do people search for online reviews after receiving personal advice? Does an incentive replace behavior that customers previously performed without compensation?
The main value of Pauwels and Aksehirli’s work is not a universal prescription. Their matrix reveals a blind spot: online conversation is not the entire market, and organic and paid WOM operate through different mechanisms.
A brand does not need to choose between “honest recommendations” and “paid influencers.” It needs to create an experience worth recommending, use paid WOM transparently and appropriately, and learn to measure conversations that leave no digital trace.
Sometimes the most important marketing contact does not happen on Instagram, TikTok, or Google. It sounds like a simple sentence from a trusted person: “I’ve used it. Here is what you should know.”
Read the complete study: Koen Pauwels and Zeynep Aksehirli, “Free and paid word-of-mouth from physical to digital, AI and beyond”, Journal of the Academy of Marketing Science, April 25, 2026. The article is open access.
Say hello to us!
A leading global agency in Clutch's top-15, we've been mastering the digital space since 2004. With 9000+ projects delivered in 65 countries, our expertise is unparalleled.
Let's conquer challenges together!
performance_marketing_engineers/
performance_marketing_engineers/
performance_marketing_engineers/
performance_marketing_engineers/
performance_marketing_engineers/
performance_marketing_engineers/
performance_marketing_engineers/
performance_marketing_engineers/
