How to Advertise on X (Twitter): Formats, Reach & What Buyers Should Know

Most platform guides answer one question: how do I buy this. For X, formerly Twitter, that is the second question. The first is what you are actually buying. The company stopped reporting as a listed entity after the October 2022 acquisition, so the reliable third-party numbers that exist for other platforms mostly do not exist here. What remains is a mix of X’s own self-reported planning-tool figures, third-party spend estimates, and a public record of litigation, regulatory findings and product change. This guide keeps those categories apart, because conflating them is how buyers end up with a plan built on a number nobody stands behind. It names no agencies.

What X reaches, and how confident you can be in the figure

X’s own advertising tools reported a global addressable audience of 586 million in January 2025, according to DataReportal’s compilation of platform planning-tool data. That is a self-reported figure. DataReportal itself cautions that ad-reach numbers are not a proxy for monthly active users, that they may include duplicate and inauthentic accounts as well as legitimate non-human accounts such as businesses and pets, and that X’s published reach data is liable to significant short-term fluctuation. Use the figures for what they are good for: relative comparison between markets, measured the same way, at the same time.

On that relative basis, X’s largest reported audiences in January 2025 ran: United States 104 million, Japan 70.9 million, Indonesia 25.2 million, India 24.1 million, United Kingdom 22.9 million, Germany 21.6 million, Turkey 19.7 million, Mexico 16.9 million, Brazil 16.0 million, Saudi Arabia 15.7 million. Two of those are far larger than population share would predict.

Japan is the outlier, and it is the reason to read this page

If there is one internationally useful fact about X, it is Japan. DataReportal’s Digital 2026 report puts X’s reported ad reach there at 71.2 million people: 57.9 percent of the total population, 64.0 percent of the eligible 13-plus audience, 63.9 percent of adults and 66.6 percent of internet users, up 1.12 million or 1.6 percent between the end of 2024 and late 2025.

Set that against the United States in the same report: 99.0 million, a bigger absolute audience but only 28.5 percent of the population and 34.3 percent of adults, down 3.79 million or 3.7 percent year on year. Japan has about a third of the US population and nearly three quarters of its X audience.

In Japanese planning terms that makes X a mainstream reach medium, not a specialist one. In the platform ranking for Japan it sits second on reported ad reach behind LINE at 99.0 million, ahead of Instagram at 63.2 million and far ahead of Facebook at 16.5 million, with YouTube at 78.5 million. A Japan social plan that skips X because a global template deprioritised it is leaving a majority-of-adults channel on the table. Our Japan market guide covers the wider mix, the LINE advertising guide covers the other Japanese pillar, and cost context sits on the X benchmarks for Japan page.

Saudi Arabia is the other market where X historically over-indexes, though the recent trend is down: DataReportal reports 15.0 million reported ad reach in late 2025, 43.1 percent of the population and 57.2 percent of adults, but a fall of 1.10 million, or 6.8 percent, year on year. See the Saudi Arabia market guide and X benchmarks for Saudi Arabia. For the two largest English-language markets, see the United States guide and the United Kingdom guide, with cost context on X benchmarks for the US and X benchmarks for the UK.

The advertiser-confidence position, sourced

This is the part most competing guides skip or editorialise. The documented record, dated:

The spend collapse. Guideline data cited by TechCrunch found 89 percent of X’s US ad dollars eroded between Q3 2022 and Q3 2024. Reuters reported in early 2023 that 14 of the top 30 advertisers had halted spending entirely and four had cut theirs by 92 to 98.7 percent. The New York Times reported US ad revenue down 59 percent year on year over five weeks in April and May 2023.

The partial recovery. The same reporting records a rebound: 65 percent of advertisers had returned by June 2024 according to internal recordings reported by the New York Times, and Guideline measured US ad spending up 62 percent year on year in the first half of 2025. Then-CEO Linda Yaccarino claimed in May 2025 that 96 percent of advertisers had returned, a company claim rather than a measured figure. She left the role in July 2025, after X was acquired by Musk’s xAI in an all-stock deal announced in March 2025.

The scale it recovered to. eMarketer estimated X’s worldwide ad revenue at 2.26 billion dollars in 2025 and 2.46 billion in 2026, which as TechCrunch noted is roughly half the 2021 business. WARC has published materially more pessimistic estimates. These are estimates, not audited figures, and they differ because there is no public filing to anchor them.

The litigation. X sued the World Federation of Advertisers and a group of large advertisers in 2024 over an alleged illegal boycott coordinated through the Global Alliance for Responsible Media. GARM was discontinued on 9 August 2024. A federal judge in Dallas dismissed X’s antitrust claim on 26 March 2026, finding X had not shown antitrust harm. On 29 July 2026 X and the WFA settled, the WFA agreeing not to form or restart GARM or a similar initiative; terms were not disclosed.

The regulatory finding. On 5 December 2025 the European Commission fined X 120 million euros, the first non-compliance decision under the Digital Services Act. The findings covered deceptive design in the blue checkmark, deficiencies in X’s advertising repository including missing content, topic and paying-entity information, and inadequate researcher data access. X has challenged the decision.

The buyer’s takeaway is neither reassurance nor alarm. The industry-wide brand-safety governance layer that used to sit between advertisers and platforms no longer exists in the form it did, and X’s own transparency tooling has been found deficient by a regulator. Both argue for buying with independent verification attached rather than on platform-reported assurance alone.

Ad formats

X documents its formats in its own creative specifications, and flags that not all of them are open to self-serve advertisers.

Promoted Ads are the core unit and look like native posts marked “Promoted”. They come as Text Ads (280 characters, reduced by 23 per link), Image Ads (PNG or JPEG, 5MB maximum, ratios including 1:1, 1.91:1, 16:9, 9:16 and newly supported 4:5 and 2:3), Video Ads (MP4 or MOV, 1GB maximum, 15 seconds or less recommended, up to 2 minutes 20 seconds supported), Vertical Video Ads (full-screen sound-on in the Immersive Media Viewer, 9:16 recommended, up to 1080 x 1920) and Carousel Ads (2 to 6 swipeable images or videos). Conversation buttons and polls layer on top as standard features.

X Amplify covers video adjacency. Amplify Pre-roll runs ahead of in-feed video from what X describes as over 200 premium publishers including TV networks, sports leagues and news outlets. Amplify Sponsorships are negotiated one-to-one packages with a single publisher. X Live supports branded livestreams with an event page and RTMP ingest up to 1920 x 1080.

Takeovers are the reach-at-scale units. Timeline Takeover places your ad as the first ad a user sees on opening X that day. Spotlight Takeover places immersive creative in the Explore tab with a 20-character trend name and optional 30-character description, supported by three to six companion Promoted Ads. X publishes no price for either.

Dynamic Product Ads and Branded Notifications round out the set. Branded Notifications run through a third-party automation platform, require a verified X account, and are open to managed advertisers in the US and Canada and to global advertisers working with X Next.

Objectives, targeting and account setup

X campaigns are objective based, and X states you are billed only for actions aligned with the objective selected. The documented objectives are Reach; Video views, Website traffic, Pre-roll views, Engagement and App installs; and App re-engagements and Website conversions.

Targeting, per X’s help centre, spans demographics (geography, language, gender, age), device targeting (operating system, device, carrier), and audience features: optimised targeting, interests, keywords, follower look-alikes, and retargeting of people who saw or engaged with past posts. Custom Audiences cover uploaded lists and website activity. Keyword targeting is the distinctive one: it reaches people who have posted or engaged with posts containing your terms, which is closer to intent than most social targeting gets, and it is the main reason X survives on plans for launches and live events. Two constraints apply before you build an audience: X restricts targeting and modifies prediction models for housing, lending and credit advertising in the US and Canada, with advertisers required to self-identify into that role, and the EU’s Digital Services Act constrains profiling-based targeting for very large platforms.

Account setup is self-serve and light: an active credit or debit card, with a five dollar pre-authorisation to validate it, and X accepts payment in the official currencies of all countries where its self-service product is available. The constraint is a ceiling, not a floor: there is no published minimum spend, but there is a credit limit, and X directs advertisers expecting to spend 5,000 dollars a month or more to apply for an insertion order to avoid it. Premium formats route through X’s sales organisation regardless of spend.

Measurement, attribution and the data-access problem

On-platform measurement is conventional: a campaign dashboard, CSV exports, the X Pixel with an Events Manager, a Conversions API deployable through tag managers, a Pixel Helper extension, AppsFlyer for app attribution, Campaign Manager 360 integration, and X’s own guidance on third-party discrepancies. If you run a content security policy, X documents that ads-twitter.com, ads-api.twitter.com and analytics.twitter.com must be allowlisted in the img-src and connect-src directives for the pixel to report correctly.

The harder problem is off-platform. X’s API is now priced pay-per-usage with no free tier: X’s developer documentation lists 0.005 dollars per post read, 0.010 dollars per user read, 0.015 dollars per post created and 0.200 dollars for a post containing a URL, with a three million post-read cap per monthly cycle on pay-per-usage plans. Social listening, share-of-voice tracking, organic-versus-paid attribution and independent verification of the environment around your ads all became materially more expensive at scale, and several tools that once offered X coverage cheaply no longer do. Layer on the Commission’s December 2025 finding that X’s ad repository lacked content, topic and paying-entity information, and independent visibility into X’s advertising environment is weaker than on comparably sized platforms.

Markets where X may not be buyable

X’s availability is politically contingent in a way that most global platforms’ is not, so verify it market by market rather than assuming. The best-documented recent case is Brazil. The Supreme Federal Court suspended X nationwide from 30 August 2024 for repeated non-compliance with its decisions, and on 8 October 2024 Justice Alexandre de Moraes authorised the immediate return of service and directed the telecoms regulator Anatel to restore it. Per the court, X had by then blocked the profiles it was ordered to block, appointed a legal representative in Brazil as foreign companies there must, and paid outstanding fines totalling 28.6 million reais. For the Brazilian market as it stands now, see the Brazil market guide.

Other markets have blocked or restricted X in the same period, but we could not source the current status of each to a court, regulator or wire-grade publication, so this page publishes no roster. Confirm in-market access yourself before fixing a launch date, and budget for substitution.

What it costs, and why one number would mislead

X publishes no rate card for auction inventory, and there is no dated, citable public CPM for a given market that we would stand behind. Any guide offering a single global X CPM is guessing, usually from a small and undisclosed account sample.

The mechanism is more useful than a fabricated average. Auction prices move with demand for a given audience, and because X’s advertiser base contracted sharply and then partially recovered, auction density varies far more by market and moment than on platforms with a stable advertiser mix. That cuts both ways: thin competition can make reach cheap, and it can also mean your ad sits among fewer comparable advertisers than your brand-safety policy assumes. Takeovers and Amplify sponsorships are negotiated per deal with X’s sales teams, so a first-time foreign advertiser and an established one pay different prices for the same unit.

For neutral cost context, start with the X benchmark hub, the benchmark database and the methodology. For regional networks, see the platform guides index.