X’s in-stream shopping prompts: how a “Get Starlink” test blurs ads and conversation
X tests inline shopping prompts placing product buttons under posts — an experiment that blends conversation and commerce and raises relevance and disclosure concerns.
A new ad that doesn’t look like an ad — what X is testing and why it matters
X has begun experimenting with a compact but consequential change to how advertising appears in its feed: contextual product prompts that sit directly beneath ordinary user posts. The early example that drew attention showed a “Get Starlink” button under a user’s post saying that Starlink’s satellite service works well in Portugal. X head of product Nikita Bier confirmed the experiment and characterized the effort as an attempt to create “an ad product that isn’t an ad.”
The experiment matters because it shifts the boundary between commercial placements and organic conversation. For X, the timing aligns with two business pressures noted by industry coverage: a search for improved monetization and a need to make advertising feel less disruptive to users. Social Media Today projected X’s revenue for 2025 at roughly $2.9 billion, about 10% higher than 2024, but still short of the platform’s pre-acquisition levels; at the same time, active EU users declined from just over 76 million to 64.8 million in the second half of 2025. Those dynamics create an incentive to extract more value from existing engagement while trying to avoid turning users off with overt advertising.
How the in-line shopping prompt operates in practice
At its simplest, the format inserts a product-focused call-to-action directly beneath a post that mentions a brand or service. The example publicly surfaced as a “Get Starlink” button positioned under a user comment praising Starlink in Portugal; unlike a promoted post, the recommendation appeared visually as part of the conversation rather than as a discrete ad unit.
Operationally, the feature represents a contextual, in-stream placement model: detection of the brand mention triggers a prompt placement tied to that specific piece of content, rather than to a user’s profile or a paid slot on a publisher surface. From an advertiser’s point of view, the appeal is straightforward — a lower-friction placement that can appear adjacent to authentic user endorsements and that therefore may generate higher engagement than conventional display or promoted posts.
The mechanics that are visible to end users are minimal — a button or short prompt that invites a follow-up action — but the more interesting design choices happen behind the scenes: what mentions qualify for prompts, how the prompt is rendered so it does not confuse the author or other readers, whether the author can opt out or disclose a paid relationship, and how clicks or conversions are routed and attributed. X has begun addressing disclosure by rolling out creator-focused ad updates like Paid Partnership labels, which provide a more structured way to indicate sponsored content without relying on hashtag disclosure; the in-line prompt experiment appears to sit next to these disclosure efforts rather than replace them.
Signal detection and the technical challenge of relevance
A central technical requirement for this format is precise content understanding. The system must determine three things reliably: which entity is being referenced, whether the mention is commercially relevant, and whether the sentiment is supportive enough to justify surfacing a product prompt.
That requirement implies a stack of capabilities: entity recognition to map colloquial mentions to specific brands or SKUs; sentiment analysis to identify positive mentions versus sarcasm, criticism, or neutrality; and contextual analysis to filter out off-topic or ambiguous references. The content that accompanied X’s test explicitly notes this need: prompts only make sense if “the system will need to identify when a mention is genuinely positive and commercially relevant.” If those signals are noisy, prompts risk appearing under posts that are neutral or critical, which would make the placement feel intrusive and could damage both advertiser effectiveness and user trust.
Operational implementations typically balance automated classification with conservative thresholds and human review loops in early rollouts. For example, a safe-launch strategy might require high-confidence positive sentiment and a clear brand entity match before showing a prompt, while marking lower-confidence cases for manual curation or exclusion. The alternative—aggressively broad placement—would likely increase impressions but also increase false positives and user backlash.
Which teams and organizations will find this format useful
Several buyer profiles will be interested in this placement if it scales reliably. Brand advertisers looking for contextual endorsement opportunities will value prompts that appear adjacent to genuine user praise because those placements can shorten the path from social proof to action. Performance marketers interested in in-stream conversion will be attracted to the tight context-to-click pipeline: a positive post followed immediately by a call-to-action reduces friction in the buyer journey.
Creator managers and partnerships teams are also implicated. X’s recent Paid Partnership labels suggest the platform is trying to give creators built-in disclosure options; combining those labels with native prompts could enable creators to participate in commerce programs more directly, either by opting into commerce placements or by using prompts as an extra monetization channel tied to their posts.
On the publisher and product side, ad ops and trust-and-safety teams will bear the operational burden of tuning contextual rules, auditing placements, and handling brand safety. Legal and compliance teams, especially for companies operating in heavily regulated markets like the EU, may need to evaluate labeling practices and how transparent the prompts are about being commercial in nature.
How this differs from promoted posts and traditional ad units
Conceptually, the in-line prompt diverges from traditional promoted posts in three key ways: placement semantics, perceived intent, and signal sourcing.
-
Placement semantics: Promoted posts are usually surfaced as paid content with a clear ad wrapper, bid into inventory like timelines or search results, and are managed through the advertising interface. The in-line prompt is embedded beneath organic posts, creating a contextual adjacency rather than displacing organic content with a paid substitute.
-
Perceived intent: Promoted posts explicitly identify themselves as advertising. The in-line prompt’s design intent is to reduce that obvious separation; to borrow X’s framing, it’s “an ad product that isn’t an ad.” That subtlety can increase click-through when the audience perceives a recommendation as authentic, but it also elevates the need for strong disclosure policies and detection accuracy.
- Signal sourcing: Traditional ads rely on targeting signals derived from user profiles, behavior, and advertiser parameters. In contrast, this format is primarily driven by content-level signals — the text of a single post and its sentiment — which changes the ad decisioning problem from audience targeting to content understanding.
From an operational standpoint, these structural differences affect reporting windows and attribution models. A content-triggered prompt will lend itself to event-based attribution tied to post impressions and immediate clicks, whereas promoted posts can be measured against longer-ranging audience exposure and campaign-level lift. The provided reporting on X’s move toward creator-focused features suggests the company is simultaneously investing in tools that make disclosure and measurement more granular, aligning ad placement with creator attribution needs.
Practical limitations and failure modes
The format’s major vulnerability is misalignment between the prompt and the post’s intent. Sarcastic posts, posts mentioning a brand only in neutral or negative contexts, or posts discussing competitors can all generate false positives if entity and sentiment detection are imperfect. An intrusive or inappropriate prompt can degrade user goodwill faster than a clearly labeled promoted post.
Another operational constraint is the transparency of the placement. If the visual design intentionally blends prompts into the conversation, users may feel misled unless disclosure mechanisms are deliberate and visible. That concern ties into regulatory and reputational risks: X’s algorithmic choices have attracted scrutiny, particularly in Europe, and blending ads and organic posts in ways that obscure promotional intent could invite additional attention.
Finally, the format requires careful calibration of advertiser expectations. Because it leverages organic conversation, performance will be uneven and dependent on user behavior in specific locales and communities. Brands that expect consistent scale might find the approach episodic; those that seek high-context, high-conversion placements may accept variability in exchange for better post-level alignment.
Measuring success: operational KPIs and experiment design
Evaluating the effectiveness of in-line prompts will require a set of content-aware metrics in addition to standard ad KPIs. Basic metrics such as impressions, click-through rate, and conversion remain relevant, but they must be measured in the context of signal quality and post intent.
Suggested operational KPIs include:
- True positive rate for matches (how often a prompt appeared under a genuinely positive and commercially relevant mention).
- False positive incidence (prompts appearing under neutral, sarcastic, or critical posts).
- Conversion per triggered prompt (clicks leading to downstream actions relative to impressions).
- Creator opt-in or opt-out rate where creators have control.
- User engagement deltas on posts with prompts versus without them, to monitor whether prompts suppress replies or organic sharing.
A robust experiment framework should use randomized controlled trials where posts that meet a high-confidence threshold are split between prompt-enabled and control groups. Manual spot checks of lower-confidence matches can help refine classifiers before broader rollout. Attribution design must account for the fact that the prompt shares context with organic social signals; multi-touch and last-click models may both understate or overstate the prompt’s causal effect, so lift testing will be necessary for claims about incremental revenue.
Where the format scales well and where it won’t
The format is likely to perform best in product categories where public, short-form praise is common and where the path to purchase is short: travel bookings, consumer hardware, digital services with straightforward signup flows, and downloadable apps are examples where a single-click or single-visit action converts well. It also favors advertisers who value contextually driven micro-moments over mass reach.
It is less well suited to high-risk categories (financial services, pharmaceuticals) where regulatory disclosure and context matter deeply; to complex B2B purchases that require sustained nurturing; and to conversations that occur in ambiguous or sarcastic tones. In those contexts, the cost of false positives or misattribution can outweigh the marginal conversion lift.
Implications for ad disclosure and creator economics
X’s experiment comes alongside platform changes intended to make commercial relationships more visible to viewers and simpler for creators to report — notably the Paid Partnership labels that remove reliance on ad-hoc hashtags. This suggests a two-track effort: enable more seamless commerce placements while simultaneously tightening disclosure tools so creators and brands can signal sponsorship without awkward manual labeling.
The tension is intrinsic: blending prompts into posts increases the likelihood that users will treat endorsements as authentic, which can improve conversion. But the same blending risks blurring accountability and disclosure, especially when creators or ordinary users are unaware that a prompt has been appended to their post. Operational governance — opt-in settings, visible labels, and predictable rules for when prompts appear — will determine whether the format becomes a scalable commerce channel or a trust liability.
What the test reveals about X’s broader strategy
The in-line prompt experiment is evidence that X is pursuing more seamless commerce pathways as part of its broader monetization push. The company’s revenue trajectory — a modest bounce toward $2.9 billion in 2025 per Social Media Today, yet still below historical highs — and a declining user base in some regions help explain the impetus to mine more value from conversational moments. By placing commercial prompts directly in the flow of posts, X is attempting to compress the path from social proof to commercial action while experimenting with ad units that feel less like interruptions.
Success will depend not only on improved classifier performance and rigorous experiment design but also on governance and transparency. If prompts reliably appear only when a post clearly endorses a product and if creators and users can easily see and control when such prompts attach to their content, the format could become a practical option for advertisers seeking contextual placements. If detection is noisy or disclosure is opaque, the placement may provoke resistance from users, creators, or regulators.
The rollout’s early stage means much will be learned from real-world signal behavior, advertiser response, and regulatory attention. How X balances the commercial upside of seamless prompts against the operational overhead of accurate content understanding and robust disclosure will determine whether this experiment becomes a routine part of the social commerce toolkit or a short-lived experiment in ad format design.
