X Ads (Twitter Ads)

Ad Formats

Promoted Posts — an ordinary post shown to a targeted audience. The core format.

Promoted Accounts — follower acquisition, shown in Who to Follow and the timeline.

Promoted Trends — a paid placement in the trending list. High reach, high cost, day-rate pricing.

Amplify and Takeover — video placements against publisher content and first-position ownership.

Targeting

Options include keyword targeting against post content and search, follower look-alike targeting that reaches users similar to a specified account's followers, interest and conversation topic targeting, and custom audiences from your own lists.

Follower look-alike targeting is the distinctive one — targeting the followers of a competitor, a publication or an industry figure is a precise way to reach a professional or enthusiast audience.

Keyword targeting against real-time conversation is the other genuinely differentiated capability, and suits reactive and event-driven campaigns.

Where X Works

Real-time and event-driven campaigns — sports, live television, breaking news, conferences. Nothing else reaches an actively-discussing audience as directly.

Specific professional and enthusiast communities — technology, finance, crypto, politics, media and gaming all retain dense active communities.

B2B awareness among audiences that follow industry conversation, though LinkedIn generally converts better.

Where It Does Not

Broad consumer direct response, where Meta and TikTok are usually more efficient. Visual-first categories, where the format is weaker. And any campaign requiring large stable reach, given audience volatility.

Brand Safety

Content moderation and adjacency on X have been a live concern for advertisers, and several large brands have reduced or paused spend over it.

Take the question seriously rather than dismissing it either way. Use available adjacency controls, review placement reporting, and make an explicit decision about acceptable risk rather than arriving at one by default. For some brands the audience access justifies the risk; for others it does not.

Building the Campaign So the Data Is Readable

X campaigns fail diagnosis more often than they fail outright, because the structure mixes signals that should be separated.

Split by targeting type, not by creative. Keyword, follower-lookalike, interest and retargeting behave so differently on cost and intent that combining them into one ad group produces an average that describes none of them. Separate them and the losing method is obvious in days rather than weeks.

Set a conversion window you can defend. The platform's default attribution is generous, and a long post-view window on a high-reach platform will credit purchases that were happening anyway. Shorten it, then compare against your own analytics rather than accepting the platform's figure.

Give creative a real rotation. Fatigue arrives quickly on a feed with high refresh, and performance decay is frequently read as audience exhaustion when it is creative exhaustion. Change the creative before concluding the audience is spent.

Brand Safety as an Operational Position

Advertiser caution about X is about adjacency — what appears next to the ad in a live, fast-moving feed — and it is not a question you settle once.

What is actually available: keyword and conversation exclusions, which work on text you can anticipate and not on what you cannot; inventory filter settings that trade reach against conservatism; account block lists, useful and endless; and third-party verification, which reports after the fact rather than preventing.

What none of them solves is that adjacency on a real-time feed is probabilistic. The honest position is a documented risk tolerance rather than an assurance — decide in advance what an acceptable rate of poor adjacency looks like, who reviews it, and what triggers a pause.

Two practical consequences. Write the escalation path before you launch, because the moment you need it is the moment there is no time to design it. And keep spend proportionate to the reputational exposure: a channel where the tail risk is a screenshot is a channel to size deliberately rather than to scale by performance alone.

Sources

What each claim on this page rests on. Entries are typed so you can see which are primary.