Read the SaaS pricing page like a contract: 7 things to check before you subscribe

Radim Václavík — Thu, 29 Oct 2026 08:05:50 +0000

Pricing pages are designed to get you to the "Start free trial" button, not to tell you what you'll actually pay in month seven. For developer and AI tools especially, the headline number is often the least important thing on the page.

Here are seven things worth checking before you put a card in. They apply to newsletter platforms, site builders, AI writing tools, hosting, and pretty much anything billed monthly.

1. What is the unit you're actually paying for?

Every plan is priced on some unit: seats, subscribers, credits, words, API calls, builds, bandwidth, pageviews. Find it, then ask two questions:

What will that number be for me in 12 months, not today?
What happens when I cross the limit: a hard stop, an automatic upgrade, or overage charges?

A plan that's cheap at 500 subscribers or 10k words a month can be expensive at 5,000 or 100k. Run the numbers for where you expect to be, not where you are.

2. Monthly vs. annual, and what the discount really costs

Annual plans are often 15–20% cheaper, and that's a fair deal if you're sure. But the discount is paid for with flexibility. Check:

Is the annual plan refundable, partially or at all?
Can you downgrade mid-term, or only at renewal?
Does the price lock for the year, or can it change?

A good rule: pay monthly for the first two or three months, then switch to annual once you know you'll keep it.

3. Credits that expire

AI tools love credits. Look for whether unused credits roll over or reset each billing cycle, and whether "unlimited" has a fair-use clause in the terms. "Unlimited" with a fair-use cap is just a limit you can't see.

4. The feature that's one tier up

Scroll through the comparison table and look for the feature you'll need, not the one you want today. Common ones that get gated:

custom domains
removing the vendor's branding
API access or webhooks
team members or roles
export in a usable format
SSO or audit logs (if it's for work)

If the one thing you need lives on the plan that costs three times as much, that's the real price.

5. How you leave

Before you start, find out how you'd stop:

Is cancellation self-serve, or do you have to email someone?
What happens to your data after you cancel, and for how long can you still export it?
Does the export include everything (content, subscribers, history), in a standard format like CSV, Markdown or JSON?

If the help docs don't answer this, ask support during the trial. How they answer tells you a lot.

6. Trial terms

Check whether the trial needs a card, whether it converts to paid automatically, and on which day. Put a calendar reminder two days before. It's not paranoia, it's just how trials work.

7. Taxes and currency

If you're outside the US, the listed price is often before VAT, and the charge may be in USD. For a small monthly tool that's a few euros of difference, but it adds up across a stack of five or six subscriptions. If you're buying for a business, check whether you can add a VAT ID so the invoice is correct.

A quick worksheet

Copy this into a note for each tool you're comparing:

Tool:
Billing unit:                     (seats / subs / credits / ...)
My volume now / in 12 months:     __ / __
Price now / in 12 months:         __ / __
Over-limit behaviour:             hard stop / auto-upgrade / overage
Annual refundable?                yes / partial / no
Credits roll over?                yes / no / n.a.
Needed feature on which tier:     __
Self-serve cancel?                yes / no
Full export format:               __
Trial auto-converts on:           __ (date)
Price incl. VAT, my currency:     __

Two tools with the same headline price can end up very different on this sheet, and it takes about ten minutes per tool to fill in.

What's the pricing gotcha that caught you out? Drop it in the comments so others can add it to their list.

Don't pay for an AI writing tool until you've run these 6 tests on the free trial

Radim Václavík — Wed, 28 Oct 2026 08:04:46 +0000

AI writing tools all look the same on their landing pages: a text box, a "Generate" button, and a testimonial from someone who 10x'd their output. The differences only show up once you use them on your own material. Free trials are short, so it helps to walk in with a test plan instead of just poking around.

Here are six tests worth running on any tool before you put a card in, whether that's Writesonic, Jasper, an SEO-focused editor like Surfer, or a plain chat model you prompt yourself.

1. Feed it something you already wrote

Take a post or README you've already published and ask the tool to rewrite the intro or summarise the whole thing.

What to check:

Did it keep your technical facts correct, or did it "smooth" them into something wrong?
Did it keep code, commands and version numbers exactly as they were?
Does the result still sound like a person, or did it add filler like "In today's fast-paced world"?

Using your own content gives you an answer key. Generic prompts don't.

2. Ask it about something niche you know well

Pick a topic where you know the details cold: a specific library, a config flag, a bug you once spent a day on. Ask for a short explainer.

Count the errors. A tool that's confidently wrong about the thing you know is going to be confidently wrong about the things you don't, and you won't catch those.

3. Test the boring jobs, not the flashy ones

Most of the real time savings in technical writing come from unglamorous tasks:

meta descriptions and post excerpts
alt text for screenshots
turning a long post into a 3-bullet summary
changelog entries from a list of commits
first drafts of FAQ sections

Run five of these in a row. If the output needs only light edits, the tool is earning its keep. If you're rewriting every one, it isn't.

4. Check what happens to your data

Before you paste anything from work, read the tool's privacy page and terms. Look for:

whether your inputs are used to train models, and whether you can opt out
how long prompts and outputs are stored
whether there's a team or business plan with different data terms

If you can't find clear answers in a few minutes, treat that as an answer.

5. Try to leave

Export something. Copy a full document out, download your history if that's offered, and see whether formatting survives. If the tool stores your drafts in its own editor, make sure you can get them out in Markdown or at least clean HTML.

It's much easier to check this on day one than after six months of drafts live inside one app.

6. Do the per-post cost math

Take the monthly price and divide it by the number of posts or docs you realistically ship per month. Not the number you hope to ship, the number you shipped last month.

If you publish two posts a month, a mid-tier plan can easily cost more per post than an hour of your own editing time. If you publish twenty pieces of content, the math looks very different. Also check the word or credit limits on the plan against your actual volume, since that's where "cheap" plans tend to pinch.

A simple scorecard

Copy this into a note and fill it in for each tool you try:

Tool:
Trial length:
1. Rewrote my own content accurately?      yes / partly / no
2. Errors on a niche topic I know:          __
3. Boring jobs needing only light edits:    __ / 5
4. Data training opt-out available?         yes / no / unclear
5. Clean export (Markdown/HTML)?            yes / no
6. Cost per piece I actually publish:       __
Keep it?                                    yes / no

Running the same scorecard on two or three tools makes the choice pretty obvious, and it keeps you from paying for a subscription because the demo looked good.

What's on your checklist that I missed? Drop it in the comments.

DEV Community: Radim Václavík