X Subscription Grok price: worth it for CV writing?

What does the new X Subscription actually cost?

SpaceXSI — the company formerly known as xAI — folded X Premium, Grok, Grok Bot and the Cursor coding tool into a single plan on 12 October 2026, priced across five tiers: Free, Plus at $8 a month, Premium at $30, Super at $100 and Ultra at $200. Annual billing trims that to $80 a year for Plus, $300 for Premium and $1,000 for Super. The pitch is bundling economics. SuperGrok, Cursor Pro and X Premium+ bought separately used to run roughly $90 a month, and the $30 Premium tier now covers all three. If you've been paying for Grok mainly to polish bullet points, that reshuffle quietly changes what you're buying.

The clever part is the shared usage pool. Instead of separate allowances, paid plans draw AI usage from one bucket that spans the Grok apps, the Grok tab on X and Cursor, so credits you don't spend on code can go to writing instead. Plus at $8 gets you Grok chat, voice, Imagine, Build, the Grok API, an X checkmark and Cursor limited to Grok models — but not Grok Bot, which now starts at the $30 Premium tier. Premium adds roughly four times the Plus allowance and access to the newest models. Super and Ultra scale that to about fifteen and fifty times, with Ultra throwing in early access to new Grok releases.

Two details matter if you're already paying. Existing SuperGrok, Cursor and X Premium subscribers aren't forced to move; switching is done through billing settings, it's rolling out gradually, and people who switch get a prorated refund. The standalone $20 Cursor Pro route is closed to new sign-ups, and third-party models inside Cursor now sit behind Premium and get billed separately at per-request API rates. For developers that's a real cost shift. For a job seeker writing a CV, almost none of this changes the one question worth asking: does any of it tell you whether your CV survives automated screening?

Is X Premium worth it for job seekers at $8 a month?

At $8 a month, Plus is a reasonable buy if you already use X and want a capable chat assistant alongside it. It is not worth upgrading beyond Plus for CV work, because no tier in the ladder includes CV scoring, ATS compatibility checking or anything that reads your file the way an employer's system does. What you're paying for is conversation volume and model access. That's genuinely useful during a job hunt — drafting outreach notes, pressure-testing how you'll explain a career gap out loud, summarising a 900-word job ad into the five things the hiring manager actually cares about — but it's a writing tool, not a screening tool.

Think about how much AI you realistically burn in a search. Most people apply to somewhere between twenty and sixty roles over six to ten weeks, and each application involves maybe fifteen minutes of prompting: tailor the summary, reorder the bullets, draft a cover letter, prep three interview answers. That workload sits comfortably inside the Plus allowance. The $30 Premium tier makes sense if you want Grok Bot and the newest models, or if you also write code. Super and Ultra are built for engineers running agents all day, and buying them to write a two-page document is like renting a cargo van to move a bookshelf.

Here's the contrarian bit: the free tier plus a few dollars spent elsewhere often beats any paid chat tier for a job seeker. Free Grok handles rewriting a bullet point perfectly well — that's a short, well-defined language task, not a reasoning marathon. The expensive capability you actually lack isn't a bigger model, it's an honest external read on whether your CV is parseable, specific and competitive. Chat subscriptions are priced on compute. Hiring outcomes turn on formatting, keywords and evidence. Those two things are not the same purchase, and conflating them is how people end up at $200 a month with a CV that still breaks in extraction.

Where does Grok help with your CV, and where does it invent experience?

Grok is strong at rephrasing material you give it and weak at anything it has to guess. Hand it a flat line like "responsible for the monthly report" plus the real numbers behind it, and you'll get a tight, active bullet in seconds. Ask it to "make my CV more impressive" with no raw material, and it fills the gaps with plausible fiction. That's not a Grok flaw specifically — every general-purpose model does it, because producing confident prose is the job it was trained for. The practical fix is to change what you ask for: give it facts to compress, never blanks to fill.

The invention patterns are predictable once you know them. Metrics appear out of nowhere — "improved efficiency by 30%", "managed a team of twelve" — when you never mentioned a percentage or a headcount. Job titles get quietly upgraded from coordinator to manager. Tools you've only read about land in your skills list because they appeared in the job ad you pasted. Certifications materialise with tidy years attached. Two short contracts get merged into one continuous role, which smooths the timeline and creates a date mismatch the moment a recruiter checks LinkedIn. Every number in your CV should be traceable to a payslip, dashboard, invoice or report you could actually produce.

Then there's the register problem. Chat models default to a glossy corporate voice that experienced recruiters now spot in a line or two: stacked adjectives, "spearheaded" on every bullet, and achievements described without a single concrete figure. Using AI to write your CV isn't against the rules — the lines sit at fabricated credentials and deception, which is roughly where Claude's own usage policy draws them too. The real risk is sounding generated while being completely honest. If you've drafted with a chatbot, it's worth checking whether your CV reads as AI-written before you send it anywhere.

Grok earns its keep on language tasks and loses credibility the moment it has to supply facts you didn't give it.

Can Grok tell you whether your CV will pass an ATS?

No — and this is the single biggest gap between a chat subscription and a purpose-built check. Grok can recite ATS rules in general terms, because that advice is all over its training data. What it cannot do is show you how your specific file comes apart during extraction. When you upload a PDF, the document is converted and flattened into context: headings, section boundaries and table structure aren't reliably preserved, and multi-column layouts measurably reduce accuracy. Uploads are session-based and don't persist, so there's no stable record to compare against later either.

That flattening creates a dangerous false positive. Your two-column designer template, the one with icons for contact details and your skills in a sidebar text box, reads perfectly coherently to Grok — it reconstructs the meaning from the words it can see and tells you the CV looks good. A real parser may interleave the columns, drop the sidebar entirely, or read your phone number as part of a job title. You get a confident green light from the chatbot and silence from the employer. The model isn't lying to you; it simply never saw the failure mode, because it isn't the thing doing the parsing.

This is where a few dollars beats a tier upgrade. The free CV analysis at HRLens takes a PDF, Word file or even a photo — Hebrew OCR included — and returns an overall score out of 100 rated Weak, Decent, Strong or Excellent, five category scores covering experience, tech stack, impact and ownership, clarity and structure, and ATS compatibility, plus a visual layout analysis and your genuine strengths. Paste the job description and it also lists the skills you're missing against that ad. If you're nervous about what a file reveals, the question of what your CV actually exposes when you upload it is worth five minutes first.

Which AI subscription is the best buy for a job search?

If you're buying tools for a job search specifically, stop at the $8 Plus tier — or stay free — and spend the difference on a CV check that scores your actual file. The maths is blunt. Ultra at $200 a month is $2,400 a year of compute allowance you will never exhaust writing applications. HRLens gives you the score, the five category ratings and the layout analysis for nothing; the Full Analysis at $3.99 adds the fix list, the "why you might be rejected" read, blind spots, a hiring probability estimate and a salary range. One of those is priced for a software team. The other is priced for one application cycle.

Be fair about where each tool wins. A chat subscription is better for open-ended, conversational work: arguing with yourself about how to frame a pivot, researching a company before an interview, generating six versions of an opening line until one sounds like you. A CV analyser is better at the parts you can't judge from inside your own document — whether the file parses, whether your impact is quantified, how you score against the market. Model choice matters less than people think for this. We came to much the same conclusion testing Mistral Large 4 on CV tasks: bigger model, same blind spots.

A sequence that actually works: get the free score first so you know where you stand, fix the structural problems before you touch the wording, then rebuild the document properly rather than patching a template that's already broken — the chat-based CV builder outputs six ATS-friendly templates, exports to PDF and Word, and writes Hebrew CVs right-to-left if that's your market. Every analysis also generates rewritten, ATS-friendly versions mirroring the target job's keywords. Use Grok for the conversational layer on top. Then cancel whichever subscription you stopped opening the week after you sign an offer, because job-search tools are seasonal and your bank doesn't know that.

How should you prompt Grok so it doesn't fabricate your CV?

Give Grok constraints, not ambitions. The prompt that works looks like this: "Rewrite these three bullets using only the facts below. Do not add numbers, tools, dates or titles I haven't given you. If a bullet lacks a measurable result, mark it with [NEEDS METRIC] instead of estimating one." That last instruction does most of the heavy lifting, because it gives the model a legitimate way to say "I don't know" — which is otherwise the one thing its training discourages. You'll get a shorter, drier draft with gaps you then fill from your own records. That's the correct output.

Feed it raw material in bulk before you ask for polish. Paste your old CV, the full job ad text, your performance review notes and a plain-language dump of what you actually did last quarter. If you're pointing it at a job posting, paste the text rather than the link — LinkedIn and Indeed block automated fetching, so a URL often yields a confident summary of a page the model never read. Then ask for one section at a time. Whole-CV rewrites in a single pass are where titles drift, dates smooth over and two jobs quietly merge into one tidier-looking role.

Finally, build in a verification pass you don't skip when you're tired. Read the finished draft against your employment history line by line and ask three questions of every claim: did this happen, can I evidence it, and would I say it out loud in an interview without flinching? Anything that fails one of the three comes out. Then run the file through a scoring check, because your own eyes have now read the document eleven times and can no longer see it. A CV that's fast to write and impossible to defend is worse than the clumsy honest version you started with.

Frequently asked questions