productivity9 min read

7 Things to Know Before You Pay for AI (How I Stay Under My Limits)

AI plans are billed in tokens, not questions. Seven things that decide how fast you hit your limit — from long chats to screenshots — and the habits that keep my subscriptions lasting all month.

Shakeel AhmedFull-Stack Developer & Privacy Tools Builder
AI usage is measured in tokens, and every message re-reads the whole chat, so long conversations, big attachments, long answers, top-tier models and extras like web search all drain limits faster. Short answers, fresh chats, cropped screenshots and the right model keep a subscription lasting longer.

As per today's date (27 September 2026) — AI prices and plan limits change often, so every price below links to the official page. Check it before you decide.

I'm on Claude's Pro plan, and I haven't hit the session limit once in the last three months. That isn't because I use it lightly — it's because of a handful of habits that all come from one thing: understanding how usage is actually counted.

Most people treat an AI subscription like Netflix: pay, use, never think about it. But AI isn't billed by the hour or by the number of questions. It's billed by tokens — small pieces of words — and almost everything you do in a chat changes how many you burn. Here are the seven things I wish I'd known before paying, and the habits that keep me under my limits.

1. You pay for the answer more than the question

On the pay-as-you-go API, output (what the AI writes) costs five times as much as input (what you send). Claude's official pricing as of today:

ModelInput / 1M tokensOutput / 1M tokensGood for
Claude Haiku 4.5$1$5Quick fixes, reformatting, simple questions
Claude Sonnet 5$2$10Most everyday work and coding
Claude Opus 5.5$4$20Hard reasoning, complex code, long analysis

On a monthly plan you don't see these numbers, but long answers still eat your allowance faster. My habit: I say how long I want the answer — "one line", "3 bullets", "just the code". Most of the time that's all I needed anyway.

2. Every message re-reads the whole chat

This is the one almost nobody knows. The AI doesn't remember your conversation between messages; every time you hit send, the entire chat so far is processed again. So message 30 in a long chat costs far more than message 1, even if both are one short question:

Earlier chat the AI re-readsThis message (question + answer)

Tokens processed per message in one 30-message chatEach message costs 500 new tokens, but the re-read chat history grows by 500 every message, reaching 14,500 of the 15,000 tokens processed on message 30.0k4k8k12k16kMessage 1: 0 tokens re-read + 500 new = 500 totalMessage 2: 500 tokens re-read + 500 new = 1,000 totalMessage 3: 1,000 tokens re-read + 500 new = 1,500 totalMessage 4: 1,500 tokens re-read + 500 new = 2,000 totalMessage 5: 2,000 tokens re-read + 500 new = 2,500 totalMessage 6: 2,500 tokens re-read + 500 new = 3,000 totalMessage 7: 3,000 tokens re-read + 500 new = 3,500 totalMessage 8: 3,500 tokens re-read + 500 new = 4,000 totalMessage 9: 4,000 tokens re-read + 500 new = 4,500 totalMessage 10: 4,500 tokens re-read + 500 new = 5,000 totalMessage 11: 5,000 tokens re-read + 500 new = 5,500 totalMessage 12: 5,500 tokens re-read + 500 new = 6,000 totalMessage 13: 6,000 tokens re-read + 500 new = 6,500 totalMessage 14: 6,500 tokens re-read + 500 new = 7,000 totalMessage 15: 7,000 tokens re-read + 500 new = 7,500 totalMessage 16: 7,500 tokens re-read + 500 new = 8,000 totalMessage 17: 8,000 tokens re-read + 500 new = 8,500 totalMessage 18: 8,500 tokens re-read + 500 new = 9,000 totalMessage 19: 9,000 tokens re-read + 500 new = 9,500 totalMessage 20: 9,500 tokens re-read + 500 new = 10,000 totalMessage 21: 10,000 tokens re-read + 500 new = 10,500 totalMessage 22: 10,500 tokens re-read + 500 new = 11,000 totalMessage 23: 11,000 tokens re-read + 500 new = 11,500 totalMessage 24: 11,500 tokens re-read + 500 new = 12,000 totalMessage 25: 12,000 tokens re-read + 500 new = 12,500 totalMessage 26: 12,500 tokens re-read + 500 new = 13,000 totalMessage 27: 13,000 tokens re-read + 500 new = 13,500 totalMessage 28: 13,500 tokens re-read + 500 new = 14,000 totalMessage 29: 14,000 tokens re-read + 500 new = 14,500 totalMessage 30: 14,500 tokens re-read + 500 new = 15,000 total151015202530Message number in the same chat15,000 tokens on message 30
Illustrative model: every question is ~100 tokens and every answer ~400. Hover a bar for its numbers. The table below has the same data.
Message #Earlier chat re-readNew question + answerTotal processed
10500500
52,0005002,500
104,5005005,000
209,50050010,000
3014,50050015,000

Across the whole 30-message chat that's 232,500 tokens processed — but only 15,000 of them are new questions and answers. The other 94% is re-reading. Anthropic says the same thing in plainer words: your limit depends on the length of your current conversation.

My habit: when a chat drifts to a new topic, I start a fresh one and paste a two-line summary of what matters.

A recent example: I asked Claude to analyse a whole project — every file. Instead of doing my coding tasks in that same heavy chat, I asked it for a summary of the points that mattered, started a new chat with just that summary, and did the coding task there. Then I did the same for each task after it. Each new chat starts light instead of dragging the entire project along with every message.

3. Files, screenshots and pasted text ride along every turn

Whatever you attach becomes part of the chat, so it gets re-read with every later message too. Images are converted to tokens by size. From Anthropic's vision documentation:

Image sizeTokens (approx.)
200×200 px (a cropped button or error)64
1000×1000 px1,296
1920×1080 px (a full-screen screenshot, newest models)2,691

My habits: I crop screenshots to just the error with our image cropper, paste only the relevant section of a document (our PDF to text tool makes that easy), and check how long something is with the word counter before pasting it. For documents I reuse, I keep them in a Project — Anthropic notes that cached project content counts less against your limits than new content.

4. The most powerful model isn't always the right one

Look at the table in point 1 again: the top model costs four times the smallest one per token. Fixing grammar, renaming variables or turning notes into bullets doesn't need it.

My rule: I use Sonnet for straightforward tasks — when the bug is clear, I understand the fix, and I can tell the AI exactly what to do. If I get stuck, or realise the problem is more complex than it looked, I switch to Opus. Most of my work never needs that switch.

5. Extras quietly drain your limit

Claude's help center lists what affects usage limits: message length, attachment size, conversation length, tool use like Research and web search, model choice and effort level. Web search, deep research and high "thinking" or effort settings are brilliant for hard questions and wasteful for easy ones.

My habit: I leave web search and extended thinking off by default and switch them on only when the question actually needs them.

6. Fix the prompt instead of piling on corrections

Every "no, I meant…" adds another turn that the whole chat gets re-read for (see point 2). A clear first message with context, the format you want and the length you want is cheaper than three corrections. If the answer misses, I edit my original message rather than replying underneath it, and I don't hammer the regenerate button — each regeneration is a full new answer.

7. Watch the meter, and pick the plan for how you actually work

As of today, Claude's plans are Free, Pro at $20/month ($17/month billed yearly) and Max from $100/month for 5× or 20× Pro's usage. The usage meter tells you whether you really need to upgrade or just change habits first.

For me the answer was clear: with the habits above, the Pro plan is enough, and I haven't needed Max.

If you're a developer paying for the API directly, two more savings: batch processing costs 50% less for work that doesn't need an instant reply, and prompt caching makes repeated input much cheaper.

The short version

  1. Ask for short answers.
  2. Start a new chat when the topic changes.
  3. Paste and attach only what's needed — crop screenshots.
  4. Use the smallest model that does the job.
  5. Turn on web search, research and deep thinking only when needed.
  6. Edit your prompt instead of stacking corrections.
  7. Check your usage meter before upgrading.

Related Topics

#ai usage limits#ai tokens explained#save ai credits#claude usage limit#ai subscription tips

About Shakeel Ahmed

Full-Stack Developer & Privacy Tools Builder

Shakeel is a full-stack developer with a focus on building browser-based tools that process data 100% locally. He created SolveBar to give developers and crypto users fast, private utilities that require no account, no upload, and no trust in third-party servers.

View LinkedIn profile →