Algorithmic advertising is designed to make promotions feel more relevant to the person scrolling past them.
But relevance alone does not determine whether an advertisement is welcomed or ignored.
New research suggests that young adults respond to algorithmic in-feed advertising through at least two different psychological routes: how enjoyable the advertisement feels and how useful or credible it seems.
A study published in Scientific Reports on 22 September 2026 analysed responses from 609 Chinese young adults who had experience with social-commerce platforms and in-feed advertising.
Personalisation and incentives were positively associated with both affective and cognitive attitudes toward the advertisements. Entertainment was linked more strongly to positive feelings, while informativeness and credibility were linked more strongly to practical evaluations.
Advertising intrusiveness showed the opposite pattern. More intrusive advertisements were associated with worse emotional and cognitive evaluations.
Both positive feeling and positive cognitive evaluation were then associated with stronger purchase and sharing intentions.
The study also found that AI transparency strengthened the relationship between cognitive attitude and purchase intention, suggesting that users who understood why an advertisement had been recommended were more likely to translate a useful evaluation into willingness to buy.
In-feed advertising sits inside the content people already consume
In-feed advertisements are promotional posts placed directly among ordinary social-media content.
They appear between videos, photographs, posts from friends and other content rather than in a separate advertising space.
Algorithms decide which advertisements to show using information such as browsing behaviour, previous interactions, purchasing activity and inferred interests.
This can make advertisements more relevant, but it also creates tension.
A highly personalised advertisement may feel useful because it matches a consumer’s needs. The same level of targeting may also feel invasive if the user believes the platform knows too much about them.
The new study examined how several different advertising characteristics contribute to that balance.
The researchers separated emotional reactions from practical judgments
The researchers based their model on stimulus–organism–response theory.
In simple terms, the framework proposes that features of an advertisement act as stimuli, which influence internal psychological reactions, which then relate to behavioural intentions.
The study distinguished between two kinds of internal response.
Affective attitude captured how positively people felt about the advertisement.
Cognitive attitude captured whether people judged the advertisement as useful, worthwhile or sensible.
This distinction matters because an advertisement can be entertaining without being useful, and useful without being entertaining.
Entertainment mainly shaped how people felt
Entertainment showed a clear relationship with affective attitude.
Young adults who perceived an advertisement as more entertaining tended to report more positive feelings toward it.
However, entertainment did not significantly predict cognitive attitude.
In other words, making an advertisement enjoyable did not necessarily make people see it as more useful or informative.
This is an important distinction for marketers because creative appeal and practical value are not interchangeable.
Information and credibility worked differently
Informativeness and credibility showed the reverse pattern.
Both were positively associated with cognitive attitude, meaning that people who saw the advertisement as informative or believable were more likely to judge it favourably on practical grounds.
Neither showed a significant relationship with affective attitude.
An advertisement can therefore help a consumer evaluate a product without necessarily making the advertisement more enjoyable.
This supports the idea that emotional and rational evaluations should be treated as separate parts of advertising response rather than combined into one general attitude score.
Personalisation and incentives influenced both routes
Personalisation was positively associated with both affective and cognitive attitudes.
When advertisements felt more relevant to the individual, respondents were more likely both to enjoy them and to judge them as useful.
Incentives such as discounts, rewards or promotional offers showed a similar dual relationship.
This makes intuitive sense.
A well-targeted advertisement can reduce the effort required to find a relevant product, while a meaningful incentive can improve the perceived value of responding to it.
But these benefits appeared alongside an equally important negative factor.
Intrusiveness damaged both emotional and cognitive evaluations
Advertising intrusiveness negatively predicted both types of attitude.
When people felt that an advertisement interrupted their browsing experience or pushed itself too aggressively into the feed, they tended to feel worse about it and evaluate it less favourably.
This finding highlights a basic limitation of algorithmic targeting.
An advertisement can be relevant and personalised while still being irritating.
More targeting therefore does not automatically mean a better consumer experience.
The placement, timing and perceived interruption caused by the advertisement remain important.
Positive attitudes were linked to both buying and sharing intentions
Both affective and cognitive attitudes were positively associated with purchase intention.
People who enjoyed the advertisement more and people who judged it more positively on practical grounds were both more likely to say they intended to buy.
The same broad pattern appeared for sharing intention.
Respondents with more positive attitudes were more likely to say they would share the advertised content with others.
The findings suggest that successful in-feed advertising is not driven only by whether someone likes the advertisement or only by whether they find it useful.
Both routes appear to matter.
AI transparency strengthened one specific relationship
The study also examined whether transparency around AI-driven recommendations changed how attitudes related to buying intention.
AI transparency refers to the extent to which users understand why a particular advertisement has appeared and how the platform has used information to select it.
The researchers found that transparency significantly strengthened the relationship between cognitive attitude and purchase intention.
When people judged an advertisement as useful or worthwhile, that positive assessment was more strongly associated with willingness to buy among users who also perceived greater transparency.
Transparency did not significantly moderate the relationship between affective attitude and purchase intention.
That suggests transparency may matter more for deliberate evaluation than for simple enjoyment.
Transparency is not the same as simply adding an AI label
The finding should not be reduced to the idea that placing the words “AI recommended” beside an advertisement will increase sales.
The study measured respondents’ perceptions of transparency rather than experimentally comparing specific disclosure labels.
Understanding why an advertisement appears can involve several things: knowing that an algorithm selected it, understanding what information influenced the recommendation and having some sense of how the system operates.
Future experiments would be needed to determine which type of explanation is most useful and whether too much detail becomes confusing or counterproductive.
The study measured intentions, not actual purchases
This is one of the most important limitations.
Participants reported how willing they were to purchase or share.
The researchers did not observe actual transactions, actual sharing behaviour or real-time responses to different advertisements in a controlled experiment.
People often express intentions that do not translate into behaviour.
A person may say they would buy an advertised product but later abandon the purchase because of price, delivery costs, competing options or simple inattention.
The findings should therefore be understood as relationships with stated intention rather than evidence that any single advertising feature directly increased sales.
The design cannot prove that the advertising features caused the attitudes
The study used a cross-sectional online questionnaire.
All of the key measures were collected from participants rather than manipulated experimentally.
This means the statistical model can identify associations that fit the proposed theoretical structure, but it cannot establish causality in the same way as a randomized advertising experiment.
For example, people who already like personalised advertising may be more likely to rate the advertisements they see as relevant and to report stronger purchasing intentions.
Other unmeasured factors may also influence both attitudes and intentions.
The sample was specific to young adults in China
The 609 participants were Chinese young adults aged 18 to 35 with experience of in-feed advertising and social commerce.
That population is highly relevant to the research question because young adults are frequent users of social platforms and social-commerce features.
However, advertising norms, platform ecosystems, privacy expectations and shopping behaviour differ across countries.
The exact relationships found in the study should therefore not automatically be assumed to apply to South African consumers, older users or people using different social-media platforms.
What this means for marketers
The findings point toward a simple but demanding balance.
Algorithmic advertising works best when the consumer receives something meaningful from the targeting.
Relevance, useful information, believable claims and worthwhile incentives can improve how an advertisement is evaluated.
Entertainment can improve the emotional experience.
But the same advertising system can undermine those benefits when targeting feels intrusive or unexplained.
The practical implication is not simply to make ads more personalised.
It is to make personalisation useful enough to justify the attention and data involved.
AI transparency may become part of the value proposition
Consumers are increasingly aware that algorithms shape what appears in their feeds.
That makes transparency more than a compliance issue.
If people understand why a recommendation is being shown, they may be better able to evaluate whether it is genuinely relevant to them.
The new study suggests that this understanding may strengthen the connection between seeing practical value in an advertisement and being willing to act on it.
But the evidence does not show that transparency will always increase purchasing.
For some consumers, greater awareness of data-driven targeting could also increase privacy concerns.
The effectiveness of transparency is therefore likely to depend on how clearly and responsibly it is implemented.
The broader lesson is that relevance has a boundary
Algorithmic advertising is often justified by the idea that more relevance creates better advertising.
The new findings support part of that argument.
Personalised advertisements were associated with more positive emotional and cognitive attitudes.
But the negative effect of intrusiveness shows that relevance has a boundary.
An advertisement can know what someone wants and still arrive in a way that feels unwelcome.
For social-commerce platforms, the challenge is therefore not only predicting what a user might buy.
It is delivering that prediction in a way that feels useful, credible and understandable rather than invasive.
Source Information
Study Title: Consumer responses to algorithmic in-feed advertising: AI transparency as a moderator among young adults in social commerce
Authors: Zhan Huang, Yan Li and Depeng Du
Journal: Scientific Reports
Published: 22 September 2026
Sample: 609 Chinese young adults aged 18–35 with experience of in-feed advertising and social-commerce platforms.
Method: Cross-sectional online questionnaire study using measures of entertainment, informativeness, personalisation, intrusiveness, incentives, credibility, affective attitude, cognitive attitude, AI transparency, purchase intention and sharing intention. The proposed relationships were analysed using partial least squares structural equation modelling (PLS-SEM).
Main finding: Entertainment, personalisation and incentives were positively associated with affective attitude; informativeness, personalisation, incentives and credibility were positively associated with cognitive attitude; and intrusiveness negatively predicted both. Positive affective and cognitive attitudes were associated with stronger purchase and sharing intentions. AI transparency significantly moderated the relationship between cognitive attitude and purchase intention but not the relationship between affective attitude and purchase intention.
DOI: 10.1038/s41598-026-72748-9
