Where X actually stands in 2026

Before allocating a single hour, get the denominator right. X is private, files nothing public, and the headline user number comes from its owner.

Elon Musk has repeatedly put X at roughly 600 million monthly active users. Treat that as a company claim, not a measurement. Independent estimates land materially lower — Backlinko put monthly actives near 557 million in mid-2025, down from about 586 million a year earlier. Similarweb data reported by TechCrunch measured X mobile daily actives at 132 million in June 2025, a 15.2% decline year over year. Over the same window Threads reached 115.1 million mobile daily actives, up 127.8%.

The revenue arc is easier to trust because multiple firms track it. eMarketer put X ad revenue at approximately $2.26 billion in 2025, up 16.5% — the first growth year since the 2022 acquisition, and still roughly 35% below Twitter’s final full year as a public company. The recovery is uneven: Q2 2025 ad revenue slipped 2.2% quarter over quarter to about $707 million before Q3 climbed to roughly $752 million.

The number that should govern your decision is not user count. It is this: X accounts for roughly 0.2% of worldwide digital ad spend, while Facebook takes 14.6% and TikTok/Douyin takes 7.1%. Only 16% of social users report using X for product discovery. If you sell physical goods to a broad consumer audience, X is not where those people shop. That is not a criticism of the platform. It is a description of what it is.

Who is there: a male-skewed audience (US users run roughly 63–66% male per Statista), concentrated in the 18–34 bracket, and unusually affluent — 29% of US users report household income above $100,000. Time spent averages 30–34 minutes a day. X is a dip-in channel for news and argument, not a dwell-time entertainment surface. Pew’s finding that 10% of users create roughly 92% of posts explains most of what the timeline feels like.

The algorithm is open source. Read it.

X is the only major platform that publishes its ranking code. xAI re-released it in January 2026. This is a structural advantage nobody uses.

The For You pipeline sources candidates (roughly half in-network, half out-of-network via SimClusters), runs a transformer to predict engagement probabilities, scores by weighted sum, then filters for diversity. X has published the weights. They are not what most social managers assume.

Three consequences fall directly out of that table. First, a reply you answer is worth 150 likes. Blocking 30 to 60 minutes after publishing to answer early repliers is the single highest-leverage action available on X, and almost no brand account does it. Second, negative signals — block, mute, report — carry large negative weights, which prices outrage bait correctly for the first time. Third, dwell time rewards threads, and thread-completion was added as a signal in 2025.

Then there is the link penalty. X’s code applies a 30–50% reach reduction to posts containing external links. Since March 2025, Buffer’s analysis of 18.8 million posts found that non-Premium accounts posting links see a median engagement rate of zero. Not low. Zero. The workaround is unglamorous and it works: put the link in the first reply, not the post.

Out-of-network reach requires social proof. Someone the viewer follows must engage before your post is eligible to surface in their feed. This is why a single retweet from a respected voice in your niche unlocks distribution that 40 of your own posts cannot buy, and why the agencies worth hiring for X sell relationships, not calendars.

Organic benchmarks and what ‘good’ looks like

The bar has moved to pay-to-play. Premium is no longer a vanity purchase; it is the price of being seen.

Sprout Social puts average brand engagement on X at 0.16%. Other datasets are harsher: median engagement around 0.015% in 2025, down from 0.029% in 2024. Rival IQ’s 2025 benchmark report recorded a 48% year-over-year decline, with the three verticals most DTC brands occupy — fashion, food and beverage, health and beauty — at close to zero. Brands post roughly 70 times a month to get there.

Premium changes the math. Buffer’s reach study found regular accounts averaging under 100 impressions per post, Premium accounts near 600, and Premium+ above 1,550. That is roughly a 10x gap between free and paid, before you write a better post. At $3–$22 a month, this is the cheapest lever in your entire content operation.

Format hierarchy, in order

X is the rare platform where text beats video. Native text posts and native video both outperform; threads drive the dwell time the algorithm now rewards; images carry roughly a 2x multiplier over plain text in the published code. External video links — YouTube, Vimeo — generate zero native video-watch signal and should be treated as links, with the same penalty.

Timing is less decisive than velocity, but it still moves the needle. Sprout, Hootsuite, and SocialPilot converge on weekday mornings (roughly 8am to noon) and midday Tuesday through Thursday. First-hour engagement velocity is the ranking lever; posting when your audience is awake simply increases the chance that velocity exists.

Cadence that actually fits a $5M–$50M brand

Two to three core posts a day, weighted to weekday mornings. Roughly 80/20 value to promotion: one substantive post (an insight, a number from your own P&L, a thread), one or two engagement posts (a question, a poll, a defensible take), plus replies. Zero external links in main posts. Links go in replies with UTMs attached.

If that cadence sounds like more than a founder can sustain by hand, it is. Most operators running a serious X presence lean on a scheduling and analytics layer to handle drafting, queueing, and reply tracking — a tool for growing on X takes the mechanical work off the calendar so the founder spends their hour on replies rather than logistics. That distinction matters, because replies are the +75 signal and scheduling is not.

Founder account vs. brand account

The brand account posts into a void. The founder account compounds. This is not a style preference; it is what the ranking weights produce.

Look at the DTC operators who have built real audiences on X: Cody Plofker at Jones Road Beauty, Sean Frank at Ridge, the founders of ButcherBox, Obvi, and Curie. These are people, not logos. Jones Road, which industry sources estimate passed $160 million in revenue in 2024, did not get there through its X handle — but Plofker’s account did get the brand into every operator conversation, recruiting pipeline, and press list that mattered.

The mechanism is legible in the algorithm. Reply-driven ranking rewards accounts that can hold a conversation. A brand account cannot argue, cannot disclose a CAC number, cannot admit a launch went badly. A founder can. That is the entire delta.

The decision tree

Question one: will the founder post three to five times a week in their own voice? If yes, lead with the founder account and let the brand handle service and announcements. If no, do not force a brand account into broadcast mode. Rival IQ’s numbers say the return is close to nothing. Use X for customer service and monitoring, and put the content hours into channels where they compound.

Question two: does your category live on X? Tech, finance, media, creator tools, and the DTC-operator community all over-index. Mass-consumer visual products do not. If you sell candles, your buyers are on Instagram and your peers are on X. Know which audience you are actually courting, and check whether your KPI framework is even set up to measure the second one.

Build in public is the dominant founder playbook and it is dominant for a reason: sharing revenue, ops decisions, wins and failures produces the replies the algorithm pays for. The payoff is recruiting, partnerships, investor access, and press. It is real. It is also not last-click revenue, and if your CFO expects it to be, the program will get killed in month four.

Paid X ads: the honest math

Cheaper clicks do not fix a thinner conversion signal. Here is the case for and against.

The costs look attractive on paper. WebFX puts X CPC around $0.74 against Meta’s $1.41, and X CPM near $6.46 against Facebook’s $7.19. Hootsuite’s dataset is even lower — CPM around $2.09, median CPC $0.18, CPA $21.55. The ranges are wide and methodology-dependent, so treat all of it as directional.

Two structural problems survive the cheap CPM. X’s conversion signal set is thinner than Meta’s, so the bidding algorithm optimizes worse toward purchase. And the audience is not in a buying posture — that 16% product-discovery figure is doing a lot of work here. A cheap click into a poorly optimized funnel targeting people who came to argue about interest rates is not a bargain.

X’s own beta data on Optimized Targeting reports a 10% average lift in click-through rate and a 16% lift in conversion rate, with 92% of advertisers keeping it enabled. That is vendor-reported, so weight it accordingly — but the ad product is not neglected. In February 2026 X expanded creative aspect-ratio support to include 4:5 (1440×1800), which is a clear signal aimed at advertisers reusing Meta creative.

Formats that exist for ecommerce

Dynamic Product Ads (requires the X Pixel or Conversions API plus a catalog in X Shopping Manager), Collection Ads (a hero asset plus up to six products), Amplify pre-roll and sponsorships, and Vertical Video Ads. X takes no cut of DPA transactions.

A testing framework that will not embarrass you

Cap the initial test at something you can lose without a conversation — most SMB tests sit between $100 and $500 a month. Run retargeting of warm audiences before cold prospecting, because warm converts and cold does not. Require the Pixel and a working catalog before a dollar moves. Set the kill criterion in advance: if blended CAC underperforms your Meta baseline after a full learning cycle, stop. Do not subsidize a channel because the CPMs feel good. A CRO partner will tell you the same thing about any traffic source that converts at half your benchmark.

X commerce: Shopping Manager, Shopify, X Money

The infrastructure exists. The momentum is somewhere else. Both facts matter.

The X–Shopify partnership announced in January 2024 is still live and was re-announced in January 2025 with easier catalog uploads. There is an X app on the Shopify App Store — free, rated roughly 2.8 out of 5. X Shopping Manager, the catalog backend, is operational and globally available.

What is gone, or at least de-emphasized: the consumer-facing storefront features from the 2021–2022 era. Twitter Shops, Shop Spotlight, Product Drops, Live Shopping. The catalog now exists to feed ads, not to host a store. No merchant-adoption numbers have ever been published, which is itself an answer.

X Money slipped from a 2025 target into a phased 2026 rollout and reached a subset of US Premium+ subscribers on June 25, 2026. At launch it is consumer peer-to-peer via Visa Direct, a metal Visa debit card, and a 6% APY deposit product through Cross River Bank. There is no merchant checkout. In-app purchase for DTC brands is a roadmap item, not a capability you can plan a quarter around.

Context sharpens the picture. Meta phased out native Facebook and Instagram Shops checkout by around August 2025, pushing traffic back to merchant sites. Reddit went the other direction, reporting that Dynamic Product Ads delivered 91% higher year-over-year ROAS in Q4 2025 alongside an 8% ROAS lift from Collection Ads. X sits in between: the plumbing is installed, the investment is going to payments and Grok. If in-app checkout is central to your platform and channel strategy, X is not the surface to build on this year.

Attribution: why X always looks worse than it is

Platform dashboards systematically undercount X. If you judge the channel on its own reporting, you will kill it prematurely — possibly correctly, but for the wrong reason.

iOS 14.5 and its successors cut trackable conversions by an estimated 30–40% across paid social. X’s thinner signal set compounds it. Demand created on X surfaces as direct traffic, branded search, and dark social — a founder’s thread gets screenshotted into a Slack channel, three people search your brand name two weeks later, and your dashboard credits Google.

The four-step setup

1. Install the X Pixel and Conversions API. Fire PageView, ContentView, AddToCart, and Purchase. Do this before spending on ads, not after.

2. Add a post-purchase survey. Fairing, KnoCommerce, or Triple Whale, placed on the order-confirmation page, where response rates run 20–30%. Include a discrete “X / Twitter” option and a free-text “Other.” This becomes your source of truth.

3. Enforce UTM discipline on every link you share — which, per the link penalty above, means every link in a reply.

4. Reconcile monthly. Compare survey-reported X share against pixel-reported. Expect the survey to be higher. Watch branded and direct search for lift in the weeks after founder content spikes. If your retention team is already running post-purchase surveys, adding one option costs nothing.

The decision rule: if the survey shows X driving more than 5% of discovery, the channel has earned a real budget line. Below that, it is a founder-brand play and should be measured on recruiting, partnerships, and press — not on ROAS.

Brand safety, risk, and the Threads question

This section is where most brand teams actually make the decision, and where most of the analysis is dishonest in one direction or the other.

Marketer confidence in X collapsed and has not fully recovered. Kantar found that only 4% of marketers consider X ads brand-safe, against 39% for Google, and that a net 26% planned to cut X spend in 2025 — the largest recorded pullback from any major platform. Marketer trust in X ads fell from 22% in 2022 to 12% in 2024.

The legal chapter closed against X. Its antitrust suit against GARM and a group of advertisers was dismissed with prejudice in 2025; the court found no illegal boycott, which is a formal way of saying advertisers left on their own. GARM itself shut down in August 2024.

Consumer sentiment is the piece brand teams underweight. Attest found 17% of US consumers name X as the least trustworthy platform. Advertising on X is legible to some customers as a political statement about its owner. That risk attaches to paid brand adjacency far more than to a founder posting in their own name, which is a useful asymmetry: organic founder presence carries meaningfully less reputational exposure than a media buy.

Where the audience is going

Threads reached 400 million monthly active users as of July 2025 per Adam Mosseri, up from 350 million in April. Its mobile daily actives hit 115.1 million in June 2025, up 127.8% year over year, against X’s 132 million, down 15.2%. Bluesky announced 40 million registered users on October 31, 2025, with roughly 3.5 million daily actives — though Similarweb data reported by Forbes showed Bluesky DAU down about 40% year over year through October 2025.

Journalists, creators, and a slice of operators have partially migrated. Not all of them, and not decisively. The practical answer is to run a founder account on X, mirror the highest-performing posts to Threads at near-zero marginal cost, and check quarterly whether your specific audience has moved, the same way you would review any other channel assumption. Do not run a full second content operation on a platform still proving out.

The 90-day operating plan

Everything above collapses into a sequence. If you cannot commit to step zero, skip the rest.

Step 0: the gate

Invest if the founder will post, or your category lives on X, or you need a cheap high-visibility service channel. Skip if you sell a mass-consumer visual product with no founder voice, if your leadership is unwilling to accept reputational exposure, or if you cannot staff sub-hour public replies. Skipping is a legitimate answer and costs you almost nothing.

Days 1–30: instrument and warm up

Buy Premium on the founder account. Install the post-purchase survey with an X option. Set the founder’s cadence at two posts a day plus 30 minutes of replies, and enforce the no-links-in-main-posts rule from day one. Do not launch paid. Do not run a content calendar for the brand account beyond service responses. Baseline your customer-service first-response time.

Days 31–60: find the signal

Add threads once a week. Track which posts generate replies you actually want to answer, not which get likes. Bring public service response time under one hour during business hours — X users are the most impatient audience you have, with 53% expecting a reply within an hour and 72% for complaints. Social support runs about $1 per interaction against roughly $6 for voice, which is the least controversial number in this article.

Days 61–90: decide

Pull the post-purchase survey data. Above 5% X-attributed discovery, install the Pixel, build the catalog, and run a $100–$500 monthly retargeting test with a documented kill threshold against your Meta baseline. Below 5%, keep the founder account and the service desk, drop everything else, and stop having the meeting. Either outcome is a win, because you now have a number instead of an argument. If the answer is that X was never your channel, your hours belong in lifecycle email or organic search, where the economics are less debatable.