Here’s the inside scoop nobody’s giving you straight: most “X advertising statistics” articles ranking on Google right now are copying numbers from each other, and half of them don’t agree on basic facts like how much a click costs.
I’ve spent months running campaigns on the platform myself, not just reading about it, and I want to walk you through what’s real. If you’re still deciding whether X earns a place in your mix, our companion piece on X’s current ownership and trust data can help you make that decision. This guide is the tactical reference to come back to once you have the available formats, what they cost, how to set up a campaign without wasting your test budget, and what the Grok in-chat ad product does before you turn it on.
A note on how to use this guide: ad platform mechanics (formats, setup steps, targeting logic) tend to stay stable for a year or more, so treat those sections as durable reference material. Cost benchmarks and anything tied to a specific feature launch age faster, so we’ve dated those claims and flagged them for periodic review rather than presenting them as fixed. Check the “last reviewed” date at the bottom before you build a budget around any number in here.
The ad formats you can run right now
X’s core ad lineup tends to stay stable for long stretches, with newer formats layered on top rather than replacing older ones. Here’s the current format lineup and what each is built for.
| Format | Best for | What to know |
|---|---|---|
| Promoted Ads (text, image) | Awareness, low-cost testing | Cheapest entry point, easiest to launch fast |
| Promoted Video | Consideration, storytelling | Widely reported to outperform text-only posts, though the size of that lift varies a lot by source, so don’t take any single stat at face value |
| Amplify (pre-roll) | Association with premium content | Placed alongside the publisher and creator video |
| Live-event ads | Real-time relevance | X’s one genuine structural advantage, still tied to news cycles, sports and live events |
| Article ads | Thought leadership, longer copy | Ties into X’s long-form article publishing feature, useful for B2B and finance |
| Community-targeted ads | Niche audience precision | Runs against specific Communities rather than broad timeline placement |
The practical takeaway: if you’re testing X for the first time, start with Promoted Ads for the cheapest signal, then move budget into video once you have a baseline cost-per-result to compare against.
What ads on X cost (and why every source disagrees)
I want to be straight with you here, rather than handing you a clean number that looks authoritative but isn’t. As of this guide’s last review, published cost-per-click figures for X range from around $0.18 to nearly $2, and cost-per-thousand-impressions figures range from roughly $2 to $6, depending entirely on which tracker you read. These figures move, sometimes quickly, so check the review date at the bottom of this guide before you plan a budget around them.
Part of that is real: cost varies enormously by objective, industry and region. But part of it is that X has been a private company since 2022 and doesn’t publish standardized ad pricing, so every “benchmark” you see is a third party’s estimate, not platform-confirmed data.
One widely cited compilation attributes a median CPC of around $0.18 and a median CPM of around $2.09 to Hootsuite’s 2025 ad benchmarks, notably cheaper than the $2.53 CPM the same compilation attributes to Meta. Other trackers in the same roundup put the average CPC closer to $0.74 or $0.38. Treat any of these as a starting planning range, not a guarantee, and build your budget around your own first-week results instead.
The same caution applies to ROI claims. You will see specific figures, such as “$2.70 for every dollar spent,” attached to X advertising across several sites, often with no cited study behind them. I couldn’t trace several of these to a verifiable primary source, so I’m not going to repeat them here as fact. Establish your own baseline in the first two weeks of a campaign and compare against it, rather than budgeting against someone else’s unverified number.
Setting up a campaign step by step
- Pick one objective. Awareness, website traffic, app installs and lead generation each unlock a different bid structure. Don’t try to optimize for two at once.
- Match the format to that objective. Use the table above. Awareness campaigns do fine with static Promoted Ads; consideration and conversion campaigns need video or Article ads.
- Build targeting from the audience out, not the budget in. X’s targeting options include keywords, interests, follower lookalikes (targeting people who follow accounts in your niche), and geography. Niche B2B and finance accounts tend to perform better with follower-lookalike targeting than with broad interest targeting.
- Set a capped test budget for week one. Somewhere in the range of a few hundred dollars is enough to get a directional read without betting your quarter on a single format.
- Launch and leave it alone for 48 hours. X’s auction needs time to find its footing. Changing bids or creative in the first two days resets the learning process.
- Review cost per result against your own baseline, not a published benchmark. This is the step most teams skip, and it’s the one that protects your budget.
- Scale what works, kill what doesn’t, and refresh creative every two to three weeks. Ad fatigue sets in faster on X’s real-time feed than on more algorithmically buffered platforms.
Grok ads: what the new in-chat product actually does
In 2025, X started letting marketers pay to appear in Grok’s AI-generated responses, essentially sponsored placement inside the chatbot’s answers rather than beside them. It’s a genuinely new ad surface, and it’s worth understanding mechanically even if you’re not ready to use it.
Here’s why it’s controversial, not just novel: other AI companies have been cautious about this exact move. OpenAI’s Sam Altman has publicly called advertising inside AI chat “uniquely unsettling.” The industry commentary picked up by Forbes keeps raising concerns: once an AI’s answers can be paid for, users lose an easy way to tell an objective response from a promotional one. That’s a real reputational proximity question for your brand, separate from whether the format performs.
There’s also a brand-safety wrinkle specific to Grok worth knowing before you test this feature. In 2026, the same chatbot was at the centre of a serious controversy over the misuse of its image-generation tools, prompting regulatory investigations in multiple countries. We cover that in full in our companion piece on X’s ownership and trust data. The short version of your media plan: run this format on the smallest test budget you’re comfortable with, monitor placement adjacency closely, and loop in your legal or comms team before scaling.
Budgeting like a platform insider
Split your first quarter of X spend into three phases rather than committing it all at once.
Phase one (weeks one to two): Small, format-agnostic tests across two or three ad types with a capped budget. The goal is data, not results.
Phase two (weeks three to six): Focus the budget on whichever format produced your best cost-per-result, and start testing audience segments within it.
Phase three (week seven onward): Scale the winning combination, hold back roughly 10% to 15% of the budget for ongoing creative refresh, and reassess the whole plan against fresh cost data every four to six weeks. Given how quickly X’s ad economics have shifted over the past year, a benchmark from six months ago is nearly useless.
If a Kantar-level trust survey or a new regulatory development materially changes the brand-safety picture, that’s a reason to revisit your phase-three commitment early, not to wait for your next quarterly review.
Platform mechanics like these hold up regardless of which way the trust numbers move, so the formats, setup steps and budgeting structure in this guide are worth bookmarking rather than re-Googling every quarter. Apply them against your own first-week data, not someone else’s benchmark, and you’ll catch a bad format or a mispriced test long before it burns real budget. If you haven’t yet decided whether X deserves a place in your mix at all, start with our companion piece on X’s current ownership and trust data before you run any of this.
If you’ve already made that call, you’re ahead of most teams just by reading this instead of copying a stat from an aggregator site. Start small, measure against your own numbers and build from there.