The AI X growth tool that writes in your voice.

Xpert learns how you write from your own posts, tells you if a draft will land before you post it, and schedules it for when your audience is actually there. You approve every post and every reply.

Then $49 a month. Cancel anytime. Nothing posts without your yes.

A week inside Xpert

app.xpert.so/overviewMon 08:30

Your morning, already sorted

What's scheduled, the slot still open, and replies drafted for posts climbing in your lane, with last week's numbers underneath. It reads like a brief, not a dashboard.

Today

Monday

This week's read

AI read

Opinions pulled 3.1x your median views. Build logs pulled 0.4x.

Next week leans opinion: two takes, one story, zero changelogs. Drafts for all three are waiting in Create.

Last 7 daysFollowers12,480up +214Views381Kup +31%Engagement rate4.8%up +0.6 pts
app.xpert.so/create/postMon 09:10

Drafts that sound like you on a good day

Xpert reads your best posts before it writes a word. Same idea, your phrasing, your rhythm, your stance. Pick a voice to see the difference.

Generic draft

What most AI tools write

🚀 Excited to share my journey! Consistency is the key to growth on X. Post every day, engage with your audience, and watch your following grow! 💯 #growth #buildinpublic

In your voice · the builder

Hit 80 · likely hit

posted every day for 90 days. followers: 212 → 1,940. the daily posts didn't do it. the 6 where i admitted something broke did.

Voice fingerprint · learned from 1,204 posts

  • lowercase
  • numbers up front
  • no hashtags
  • short lines

Dropped from the generic draft: rocket emoji, “Excited to share”, 2 exclamation marks, 2 hashtags.

Sample voices. When you connect, the voice is learned from your own posts.

app.xpert.so/create/postMon 09:14

Know if it lands before you post it

The hit meter scores the draft against your own history and points at the weakest line. This one runs in your browser, so edit the text and watch it move.

129 / 280

A ten-check sketch that runs in your browser. In the app, the score comes from your own posting history and recalibrates as results come in. What X rewards

80

Likely hit

  • Fix: A stance Say what you believe, not only what happened.
  • Fix: Talks to the reader Address them, or ask them something.
app.xpert.so/calendarMon 09:20

Posted when your audience is actually there

Queue slots come from when your followers engage, not a generic best-times chart. Posts, threads and X Articles share one queue, and everything stays movable.

When your audience engages

Sample audience activity by day and two-hour block, as a share of the weekly peak
060810121416182022
Mon
Tue
Wed
Thu
Fri
Sat
Sun

Pick a block to see how active your audience is at that hour

LessMore Open slot Queued

Queue

Learned from your replies
  • Mon 10:00stopped posting progress, started posting opinions about the progress
  • Tue 08:00Open
  • Wed 12:00Open
  • Thu 18:00Open

Next up: posted every day for 90 days.

Drops into the best open slot. Nothing is posted outside a slot you chose.

app.xpert.so/inspiration/foryouWed 12:40

Replies worth writing, sent by you

Posts from people you follow that are climbing in your lane, each with a reply drafted in your voice. Nothing reaches anyone until you approve it.

  • @amelie_buildsClimbing2h

    shipped v2 of our onboarding and retention went up 18%. one screen did most of it.

    Drafted in your voice

    18% from one screen is wild. was it the first screen or the empty state? my bet is the empty state.

  • @theo_writesHot40m

    hot take: threads are dead. single posts with one sharp opinion win every time now.

    Drafted in your voice

    half agree. threads die when they're a list. the ones that still work are one argument, told in order.

  • @ken.devNew12m

    which X analytics tool do people actually pay for? mine shows views and nothing else.

    Drafted in your voice

    pay for the one that tells you why a post worked, not just that it did. views without reasons are a vanity feed.

Approvals

Waiting for you
3
Sent by you
0
Sent without your yes
0, and no setting changes that

Why replies? X's ranker gives a reply 7.6x a like's weight, and a reply between mutual follows 30x.

In July 2026 X removed 42,000 accounts that were automating replies with chatbots, per X's head of product. Xpert is built so that can never be you.

app.xpert.so/analytics/postsSun 18:00

Every week makes the next one sharper

Views, followers and engagement from your own history, broken down by what you posted. The same data retrains the writer and the hit meter.

381K this week

40K50K60K70KMonTueWedThuFriSatSun76.1K

What worked

vs your median post
  • Opinions3.1x
  • Stories1.8x
  • Threads1.2x
  • Build logs0.4x

The thin line marks your median (1x).

Recalibrated

For you, one opinion post is worth about eight build logs. Next week's drafts and hit scores lean that way.

Your voice, measured on every draft

Voice match, Voice DNA, Edit memory, Hook Lab, the pre-flight check and the weekly read come with every plan.

app.xpert.so/create/postSample account

1Voice match

Sounds like you, with a number on it

Xpert measures how you write from your own posts: length, rhythm, line breaks, casing, punctuation, emoji and the words you never use. Every draft then gets a sounds-like-you score against your closest posts, with the lines that drift marked and a one-tap fix.

  • AI phrases you never use cost points
  • Every Voice DNA trait has a switch, for posts, threads and replies
app.xpert.so/settings/voiceSample account

2Edit memory and your takes

Learns from you, not from everyone

Cut the opening or swap leverage for use, and Xpert learns the rule and uses it in every new draft. A ten-minute interview fills a bank of your stories, opinions and numbers, so drafts use your takes instead of filler. Or talk for two minutes and get a post, a thread and two reply-ready takes.

  • The rules sit in Settings, each with a remove button
  • A three-question check-in each week keeps the takes fresh
app.xpert.so/create/postSample account

3Hook Lab

Twelve ways in, ranked

Hook Lab writes twelve openings for one idea across six angles: story, contrarian, specific, question, list and lesson. They come ranked by how likely each is to stop the scroll and how much it sounds like you, and the one you pick teaches Edit memory your taste.

  • The rank weighs stopping the scroll at 65% and sounding like you at 35%, each against the best of the set

Idea: Replies grew my account from 212 to 1,940 followers in 90 days, more than posting daily did.

  1. Rank 1Specific212 to 1,940 followers in 90 days, and the daily posts did the least.Stops the scroll 90 of 100Sounds like you 100 of 100
  2. Rank 2ContrarianPosting every day is the most overrated growth advice on this app.Stops the scroll 100 of 100Sounds like you 81 of 100
  3. Rank 3ContrarianYour posts aren't what grows your account. Your replies are.Stops the scroll 93 of 100Sounds like you 90 of 100
  4. Rank 4Story90 days ago I had 212 followers and a daily posting habit that wasn't working.Stops the scroll 78 of 100Sounds like you 92 of 100
  5. Rank 5LessonMy biggest growth mistake: posting daily and replying to nobody.Stops the scroll 82 of 100Sounds like you 84 of 100
  6. Rank 6List3 things that took me from 212 to 1,940 followers. None of them is posting daily.Stops the scroll 86 of 100Sounds like you 76 of 100

Stops the scroll Sounds like youEach against the best of the set.Top pick, Specific. The real number first, then a twist on the usual advice.

app.xpert.so/create/postSample account

4Pre-flight

Checked before it goes out

One checklist before you post: voice, AI tells, mute and report risk, creator payout safety, originality, the link price, the AI-image label, and a fact check against the web when you want one. If a draft fails a check, compose asks you once before posting it.

  • X weights a mute at −58.8 and a report at −234, against 0.66 for a like, because both are rare
  • Engagement asks and copied posts cost creator payouts
app.xpert.so/analytics/postsSample account

5Weekly read

Then it tells you what worked

A day or two after each post, Xpert gives it a verdict against your own usual (outperformed, typical or underperformed) and a two-line lesson. Every Monday the weekly read says what worked, what didn't, one thing to try next, which posts are worth rewriting and how close the pre-post scores came.

  • What works for you: your own winning formats, not everyone else's

Try the phrase check on a draft of yours

Paste a draft or pick a sample. The phrases people now read as AI light up as you type, with the swaps Xpert offers. It runs in your browser, so nothing you type leaves this page.

The rules are the same ones the composer, the pre-flight check and the free AI slop detector use.

app.xpert.so/create/postDemo
174 / 800

A demo of one part of voice match. Without your posts this page can't tell how much a draft sounds like you, so it starts every draft at 100 and takes off what these phrases cost in the app. In the app the score starts from your own closest posts, and a phrase your posts use costs nothing.

Demo read

30%

Reads as AI slop

8 AI phrases took 70 points off.

  • In today's fast-paced digital landscape,In today's world

    A warm-up that says nothing. Start with the point.

  • paramountBuzzword

    Sounds impressive, carries no information.

  • FurthermoreFurthermore

    Formal glue nobody uses out loud.

  • it's important to note thatIt's important to note

    If it matters, just say it.

And 4 more, marked in the draft.

What the X algorithm actually rewards

X open-sourced the ranker behind its For You feed. For every viewer it predicts the odds of each action, multiplies each one by a published weight, and adds them up. Here are the real weights from that code, and a sandbox to see what they do to a post.

What lifts a post

What lifts a post: published weight per action, and how many times a like's weight it is
ActionWeightTimes a like's weight
Share via copy link2030× a like
Reply57.6× a like
Quote57.6× a like
Share via DM57.6× a like
Follow the author46.1× a like
Share23× a like
Repost11.5× a like
Like0.661× a like
Click0.30.5× a like
Open a link0.20.3× a like

What sinks it

What sinks it: published weight per action, and how many times a like's weight it is
ActionWeightTimes a like's weight
Report−234355× a like
Mute the author−58.889× a like
Not interested−52.27279× a like
Block the author−31.247× a like

Each group has its own scale so both stay readable. The weights multiply each viewer's predicted chance of the action, not a count of them: reports are over 1,000 times rarer than likes, so their weight has to be large to move the ranking at all.

A post's first 1,000 views

Set what those viewers do. The score uses the weights on the left.

The viewer
Weights as published.

Ranking score

62

Worth 93 likes · ranks 3 of 4

Likes are the biggest line here, and a like is the cheapest engagement X counts: a copy-link share carries 30x its weight. 1 mute took back 59 points.

How X would rank them

  1. Worth sharing+331
  2. Conversation starter+289
  3. Like magnet+62
  4. Rage bait−444

Illustrative numbers. The real ranker predicts each action's odds per viewer; here you set them.

Strangers count for less
Posts shown to people who don't follow you are scored at 0.75x, and so are your replies and reposts, even to followers.
Mutuals' replies count 4x
On an original post, a reply from someone you follow back is weighted 20 instead of 5.
Bursts get discounted
Your second post in the same feed is scored at 62.5%, your third at 44%. Space them out.
New posts get a launch window
Up to 50,000 followers, a new original post can be lifted to slot 16 of a feed for its first 2 hours or 200 home-feed views. X picks by likes per view so far.
Reading after the click counts
Since late September, time spent on a post after opening it carries a weight of 0.4. In October a like rose to 0.66 and opening a video fell to 0.02.
What Xpert does with it
Ranks your past posts with these weights to learn what works for you, and reads every draft for reply pull, shareability, follow intent and mute risk.

Weights and rules from X's open-sourced ranking code, xai-org/x-algorithm (home-mixer/params/param.rs), as published in 2026. X can change them at any time.

How the first three days go

The trial is short on purpose. Everything worth judging happens in the first sitting.

app.xpert.so/login

Start with your 𝕏 account

Xpert reads first, and only posts what you schedule or approve.

  • Your posts and their numbers
  • Your profile
  • Who you follow, for the network brain
Continue with 𝕏

Selling something? Ask it for buyers.

The assistant can see your numbers and act on them. Ask for people you can sell to and it searches 𝕏 live, qualifies each person against what you sell, and drafts the reply in your voice. Or ask it to grade your month, x-ray an account, or plan your week.

app.xpert.so/assistant Sample replay

Find me 5 new people that I can sell my product to

A replay with sample data. The leads card is live: tick, draft, save.

Everywhere you write

One voice and one set of approvals behind the app, x.com, your iPhone and the AI assistants you already use.

The extension, beside x.com.

The whole app opens in Chrome's side panel, next to the post you're reading. Its "This post" card drafts a reply in your voice and predicts how it will land, the Xpert button on every post does the same on x.com, and AI Shield badges replies that look AI-written or automated. X's reply box opens with your draft ready; you press Reply.

X only accepts replies through its API when the author mentioned you, and in July it removed 42,000 accounts that automated replies, per X's head of product.

See the extensionAdd to Chrome

Voice notes on iPhone.

Talk for two minutes and get a post, a thread and two reply-ready takes, anywhere from keep my words to a full rewrite. The Safari extension, for iPhone and Mac, is coming.

Your voice in Claude and ChatGPT.

Connect the MCP server and your assistant gets your voice guide, a draft-in-my-voice prompt and the pre-flight check, along with your drafts, queue and analytics. Script it with the CLI, or call the REST API.

Set up MCPDevelopers

Repurpose studio.

Paste a YouTube video, a podcast, a blog post or a newsletter and get a thread, five posts and an X Article in your voice. Quotes and figures are checked against the source, and nothing is copied from it word for word.

Growth knows your audience. Scale writes for a team.

Everything above comes with every plan. These come with the bigger ones.

GrowthScaleplans

Knows how your audience will take a draft

The audience simulator shows how each group of your followers would react to a draft: who replies, shares or mutes, weighted the way X ranks. It compares versions and says how sure it is. It never forecasts views.

app.xpert.so/create/postSample audience

Your draft

Replies grew my account more than posting every day did. 212 to 1,940 followers in 90 days.

How each group would react, against your first draft

  • Founders · 34% of followers

    High confidence
    up +55%

    Per 1,000 who see it: 6 replies, 5 shares, 1.0 mutes

  • Writers · 28% of followers

    High confidence
    up +22%

    Per 1,000 who see it: 8 replies, 5 shares, 0.8 mutes

  • Marketers · 22% of followers

    Medium confidence
    up +84%

    Per 1,000 who see it: 4 replies, 3 shares, 0.9 mutes

  • Quiet readers · 16% of followers

    Low confidence
    up +34%

    Per 1,000 who see it: 1 replies, 1 shares, 0.3 mutes

Across your followers

up +41%

Scored with X's published weights: a copy-link share 20, a reply 5, a repost 1, a like 0.66, a mute −58.8. The audience and its rates are made up for this demo. A comparison between versions of a draft, not a forecast.

Best-model match
Each month Xpert tests which AI writes most like you, and drafts with the winner.
Mutuals radar
The mutuals most likely to reply to you. On an original post, a mutual's reply weighs 20 instead of 5.
Smart spacing
Room between your posts: your second post in the same feed keeps about 62.5% of its score. Up to 50,000 followers, it also shows the 2-hour launch window a new original post can win.
Coach lessons per follower group
Post verdicts and weekly reads broken down by which groups of your followers responded.

Also in Growth: Autopilot, real-time keyword monitors and competitor tracking.

Scaleplan

Writes for a team

Council drafting has several top models write every draft at once. You see the best two, and why each one won.

app.xpert.so/create/postSample account
  1. Model A

    Best

    212 to 1,940 followers in 90 days. The replies did it, not the daily posts.

    Closest to your voice: the number first, short lines.

  2. Model B

    Runner-up

    Posting daily didn't grow my account. Replying to 10 bigger accounts every morning did.

    Strongest opening: a claim people want to answer.

  3. Model C

    Not picked

    Unlock your growth: consistency in replies is the key to building an engaged audience.

    Left out: two AI phrases you never use, and no number.

Canned drafts for a sample account, replayed in your browser.

Voices and a client workspace
Founder, brand and Articles voices on one account. Share a draft link for client approval and comments, with a log that marks edits made by an AI assistant apart from people's.
Auto-tuned voice
Each week Xpert tests different examples and rules in your voice prompt and keeps whatever lands closest to your real posts.

Plus the top limits on credits and accounts.

Three plans, from $49 a month

And how it stacks up against the tools creators usually compare it with, including where it doesn't win.

GrowthMost popular

Regular price $99$49/month

Billed monthly

Sale price until October 31, 2026. Join by then and keep $49 a month.

  • Drafts in your voice: posts, threads and X Articles
  • Voice match, Voice DNA and Edit memory on every draft
  • The hit meter, Hook Lab and a pre-flight check
  • A queue built on your audience's best times
  • The weekly read and a verdict on every post
  • For you replies and the Approvals inbox
  • Unlimited scheduled posts
  • Your voice in Claude and ChatGPT, plus the API and CLI
  • The Chrome extension beside x.com
Start for free

Free trial, then $49 a month. Card required, cancel anytime.

Also Pro at $49 and Scale at $199. Compare plans

Hypefury is left out: its own pricing FAQ says it no longer supports X.

Every reply and DM approved by you

Xpert
YesAlways. No setting turns it off
Typefully
NoAuto-DMs and Auto-plug send on their own
SuperX
NoAuto DM, Auto Plug and DM campaigns send on their own

Scores a draft before you post

Xpert
YesAgainst your own history
Typefully
No
SuperX
YesViral Score

Shows how much a draft sounds like you

Xpert
YesVoice match on every draft, with the lines that drift
Typefully
No
SuperX
No

Flags mute, report and payout risks before you post

Xpert
YesPre-flight check, free on every draft
Typefully
No
SuperX
No

Learns your voice from your posts

Xpert
YesPlus Edit memory, from every change you make
Typefully
YesCreator plan and up
SuperX
YesAdvanced plan and up

Reply drafts in your voice

Xpert
YesYou send every one
Typefully
No
SuperX
YesAdvanced plan and up

Best times from your own audience

Xpert
Yes
Typefully
Yes
SuperX
Yes

API and MCP server for AI agents

Xpert
YesPlus a CLI
Typefully
Yes
SuperX
YesPaid plans

Cross-posts to LinkedIn, Threads, Bluesky

Xpert
NoX only, for now
Typefully
Yes
SuperX
PartlyBluesky only

Free plan

Xpert
NoFree trial
Typefully
Yes
SuperX
NoFree trial

Checked against each product's own site and docs on 24 September 2026, and the voice match and pre-flight rows on 27 September. Plans change; if something here is out of date, tell us at contact@xpert.so.

50+ free X and LinkedIn tools. No login.

Try the writing, the scoring and the timing before you sign up.

X growth tool FAQ

Straight answers about the writer, the predictor, the algorithm, the price, and keeping your account safe.

What is the best AI tool to grow on X (Twitter)?

The best AI X growth tool is the one that grows your account without making you sound generic. Xpert trains on your own posts, scores every draft against your history with an engagement predictor, schedules at your proven best times, and drafts replies to the posts worth answering. Nothing publishes without your approval.

Will Xpert make me sound like AI?

No, that is the whole point. Xpert learns your voice from your own past posts and drafts in it: your phrasing, your rhythm, your stance. No hashtag spam, no thought-leader filler. You edit and approve everything before it ships, so nothing goes out that you would not have written.

How does Xpert know a draft sounds like me?

Voice match compares each draft with six of your own posts closest to it in topic and gives a sounds-like-you percentage, with the lines that drift marked and one-tap fixes. AI phrases you never use cost points. Your Voice DNA (length, rhythm, line breaks, casing, punctuation, emoji, openers, endings, signature words and the words you never use) sits in Settings, where every trait has a switch, and Edit memory learns from every change you make to a draft.

How does the engagement predictor work?

Every draft is scored against your own posting history, not a generic viral template. The hit meter compares the draft with what has actually worked for your audience, gives a verdict, and highlights the weakest line. It recalibrates as new results come in, so the predictions track your account, not someone else's.

What does the pre-flight check look at?

Before you post, Xpert checks the draft for voice match, AI tells, mute and report risk, creator payout safety (engagement asks and copied posts), originality, what a link costs and the AI-image label, and it can check facts against the web. X weights a mute at −58.8 and a report at −234, against 0.66 for a like: the weights scale each reader's predicted chance of the action, and these are rare, so even a small chance costs reach. If a draft fails a check, compose asks you once before posting it. The check is free; the fact check uses credits.

What does the X algorithm reward in 2026?

X's open-sourced ranker predicts what each viewer will do with a post and weights each action. A share via copy link is weighted 20, a reply, quote or share by DM 5, a follow 4, a repost 1 and a like 0.66. Negative signals weigh far more: a report is -234 and a mute -58.8. Posts shown to people who don't follow you are scored at 0.75x. Xpert ranks your past posts with these weights and scores drafts for what they reward.

Does Xpert schedule threads and X articles?

Yes. The tweet scheduler handles single posts, full threads posted as real reply chains, and long-form X Articles with AI-generated covers. Queue slots are learned from when your audience engages, and you can send anything to drafts instead of the queue.

How much does Xpert cost?

Xpert has three plans: Pro at $49 a month, Growth at $49 (the most popular; sale price for anyone who joins by October 31, 2026, normally $99 a month) and Scale at $199, each with a 3-day free trial. A card is required to start, nothing is charged during the trial, and you can cancel from Billing in one click. It includes the app, the REST API, the CLI and the MCP server.

Will Xpert get my account banned for AI content?

X does not ban AI-assisted writing, and it explicitly allows scheduled original posts. What it removes is accounts that let AI engage with other people with no human involved: in July 2026 it deleted 42,000 of them in one sweep. Xpert cannot do that. Every reply and DM to another person waits in an approval inbox until you read it and send it, and there is no setting that turns that off. You still own what you post, so check drafts for accuracy before approving them.

Is automation safe for my X account?

Xpert is built human-in-the-loop: replies and DMs are drafted by the machine and sent by you. Outbound actions wait in an approval inbox, sends are capped and spaced like a person typing, and campaigns pause the moment X pushes back. Your account is the asset, so nothing risks it on autopilot.

Can Xpert find customers for me?

Yes. Ask the assistant for people you can sell to and it searches X live for posts from people asking for what you sell, then qualifies each one against your product. You get a short list with a one-line reason each, ready to add to Contacts or to reply to in your voice.

Can AI agents use Xpert?

Yes. Xpert ships an MCP server that agents connect to over OAuth, a CLI for scripting from the terminal, and a REST API. Claude, Cursor, or your own agent can draft, schedule, and read analytics through the same brain the app uses.

What does Learning mode do with my data?

Learning mode is off by default. When you switch it on in Settings, Xpert records what you click, which pages you use, and what you write inside the app, and distils that into a private profile that tunes the assistant, the composer and your starter prompts. Raw signals are kept 60 days; the profile stays until you clear it, which you can do any time.

The feed forgets you in three days.
Xpert doesn't let it.

Start for freeConnect MCPTakes about two minutes. It reads your posts while you make coffee.