On TikTok, a Like Can Change the Ad Load—Without Proving Purchase Intent

A September 30 arXiv preprint found that liking and sharing on TikTok increased the advertising shown to automated accounts. The finding creates a commercial tension: an action that expresses interest can also change the environment in which the next ad competes. For marketers, the question becomes whether engagement signals demand, greater exposure, or a mixture of both.

The platform reads an action. The marketer reads a motive.

In “From Scroll to Sale,” Nazanin Sabri and colleagues audited more than 80,000 videos across 56 automated accounts. In the first experiment, ads represented 25.7% of videos in the scrolling condition, 30.8% with liking and 35.2% with sharing; commenting produced 25.9%. These are experimental feed observations, not population-wide TikTok averages.

TikTok’s own “The Basics of TikTok Ads” explains that viewing, searching, sharing, liking and previous ad interactions can inform advertising personalization. Its Ads Manager documentation also lists device type among targeting dimensions. Those disclosures establish that behavior and device information can matter; they do not explain every observed delivery decision.

The strategic problem is the distance between an observable action and its interpretation. A like might mean admiration for the choreography, agreement with an opinion, recognition of a joke or interest in the product. A share might be a recommendation, a warning or a private conversation starter. None of those interpretations should be silently converted into willingness to pay.

Consider a hypothetical campaign for an expensive suitcase. A beautifully staged packing film earns enthusiastic reactions. Those reactions could identify future travelers, fellow creators studying the production or people enjoying the transformation. The same visible action could describe three very different relationships to the offer. Treating them as one audience would conceal the distinction the creative needs to resolve.

More exposure is not more demand

A useful reading of the audit is that the advertising environment can respond to activity. For a brand, that creates a measurement question beyond whether one post performed well: what happens when an interested viewer encounters the next message among competing offers? The study does not answer which advertiser wins that contest.

This is why an engagement result and a commercial result need separate names. A film can succeed at making an idea memorable while leaving the purchase decision untouched. Another execution can attract fewer reactions but address a decisive objection. Comparing the two only by likes would answer a narrower question than the business intended.

The next analytical step is to specify the uncertainty. Is the audience unconvinced by the product, uncertain about delivery, worried about fit or simply not shopping? A creative concept that resolves one of those obstacles has a defensible commercial purpose. A concept that only repeats the spectacle may earn another reaction without revealing what changed.

A device is a clue, not a customer

The audit also compared phones by price, using device cost as an income proxy. The cheapest device received larger advertised percentage discounts, but the authors caution that these did not necessarily mean cheaper deals. Device price is an ambiguous signal, and the audit does not establish that identical goods had different checkout prices.

A marketer should resist filling that gap with a story about wealth. A person can keep an older phone by choice, use a work device or buy a refurbished premium model. These are plausible counterexamples, not observations from this experiment. They show why a convenient proxy requires a more modest interpretation than a verified customer fact.

There is a positioning decision here too. If a campaign leads with a discount, it asks the customer to assess savings. If it leads with durability, it asks the customer to assess value over time. Before comparing results, the business should recognize that it has changed the question presented to the audience—not merely the decoration around the same offer.

Keep the customer's decision in the measurement

The preprint is useful evidence about ad exposure, with a limited account sample and automated behavior. It is not proof of purchase intent or a sales forecast. That boundary makes the practical lesson sharper: define the decision an advertisement is intended to influence before treating a platform reaction as success.

For an awareness campaign, the question might be whether people remember the brand and its proposition. For a conversion campaign, it might be whether qualified buyers complete the purchase at an acceptable acquisition cost. For a pricing experiment, it might be whether the offer improves contribution after discounting. Each requires evidence appropriate to its own claim.

For OrionPilot, this is a useful standard for marketing direction: keep the creative hypothesis, customer obstacle and intended business outcome connected. A weekly review should preserve the difference between “people responded” and “the commercial decision improved,” so that the next campaign learns from the right question.

The most valuable response to the TikTok audit is therefore not to stop seeking attention. It is to stop asking attention to prove more than it can. Before celebrating the next surge in reactions, identify what the customer still needs to decide—and what evidence would show that the campaign helped them decide it.