What $100 buys
Check for backlinks free, then price the rest in credits
Four actions need no key and no balance. Everything else is priced per call, from $0.0001. A rough ledger beats surprise.
SEO AI Pal research team ·
Four actions that cost nothing
No key, no card, no signup for these. They are the right place for an assistant to practise.
domain
The same call the DR widget on the home page uses. Domain in, rating out.
overview
A first summary of a domain. Start every chat here.
stats
Index-level numbers, handy to confirm the endpoint is alive.
suggest
Name completion, so the assistant can confirm a domain before spending anything.
Prices per call
| Credits | Dollars | Actions |
|---|---|---|
| 0 | $0 | domain, overview, stats, suggest |
| 1 | $0.0001 | bulk_dr (up to 2,000 domains), compare (up to 20), top, dr_distribution, tld_breakdown |
| 2 | $0.0002 | refdomains, backlinks, anchors, toppages, anchor_cloud, intersect (2-10 domains) |
| 3 | $0.0003 | search, keyword |
What a hundred dollars looks like in chat
The sign-up credit is $100, or 1,000,000 credits. At 2 credits a call, that is 500,000 refdomains pulls. Nobody types that many questions.
The danger is not the price. It is a loop. An assistant told to "check all my competitors weekly" can fire dozens of calls per run, and a retry rule that never ends turns a quiet Tuesday into a long ledger.
Set three limits: one retry per call, a maximum number of calls per chat written into the prompt, and a habit of reading _credits after each reply. The wording lives in the runbook.
Example ledger (hypothetical)
One month of a solo consultant's chats
Suppose 40 chats, each with one overview (free), three refdomains pulls and two anchors pulls. That is 5 paid calls at 2 credits, so 10 credits per chat and 400 credits for the month, about $0.04. Invented numbers for illustration.
- Add one bulk_dr over 2,000 prospects: 1 credit.
- Add a monthly compare of 20 rivals: 1 credit.
- Still under five cents.
Why the cheap part is the honest part
Pricing this low is possible because the product is narrow. It reads one link graph of 2,774,625,913 domain-to-domain links, refreshed monthly, and does not crawl keywords or ranks.
Ahrefs sells API units inside subscriptions. Plans start at $129 a month for 200,000 units, with at least 50 units per request. That suits a team that lives in Ahrefs, while prepaid credit with no monthly minimum suits someone testing a prompt on a Saturday afternoon.
Mind the shape. Compare cost per answer rather than cost per plan. Fifty units for a single request is a different shape from 2 credits. See the myths page for the reason we care about receipts.
Guardrails to paste into the instructions
Money is rarely the problem. Surprise is. Write the limits where the model reads them first, and repeat them in the tool description.
Something like: at most ten paid calls per chat, one retry per failed call, and stop the moment the status is 402. Ask the assistant to print the remaining _credits at the end of every reply.
If you later move from chat to a scheduled script, port the same three limits into code. A loop in a script is faster than a loop in a chat, and the balance moves with it.
Before you hand the key to an assistant
- Test every prompt on the free actions first.
- Expose only the actions the job needs.
- Write the retry limit into the instructions.
- Keep the GPT or tool private while you test.
- Read the balance once a week from the account page.
- Top up in small steps. USDT (TRC-20) works from the account page.