What an automated source-check pipeline costs, and where it stops
If you plan to run checks on every draft your team generates, you need the unit costs and the ceilings written down. Here they are, plus a hypothetical month and a plain comparison with an Ahrefs DR checker.
Rate card and limits
| Action | Credits | Limit per call | Good for |
|---|---|---|---|
| domain, overview, stats, suggest | 0 (free) | Single domain, no key | Spot checks while reading |
| bulk_dr | 1 | Up to 2,000 domains | Scoring a draft's whole reference list |
| compare | 1 | Up to 20 domains | Choosing between candidate sources |
| top, dr_distribution, tld_breakdown | 1 | Up to 1,000 rows | Context about a publisher's footprint |
| refdomains, backlinks, anchors, toppages | 2 | Up to 1,000 rows, paging to 50,000 | Digging into one doubtful source |
| intersect (2 to 10 domains) | 2 | Up to 1,000 rows | Finding who links to several sources |
| search, keyword | 3 | Up to 1,000 rows | Anchor and title lookups |
Errors your job should handle
| Code | Meaning | What the job does |
|---|---|---|
| HTTP 401 | Key missing or wrong | Stop and alert; do not retry |
| HTTP 402 | Credit has run out | Pause the queue and notify whoever tops up |
| ok: false | The request was rejected or malformed | Log the response and skip that batch |
| Empty rows | Nothing in the index yet | Mark as unverified, send to a human |
A hypothetical month
This is an example, not a forecast. Suppose a team publishes 120 AI-assisted pieces in a month, and every piece goes through the same two-step check before an editor signs it off. Numbers are invented.
Each piece gets one bulk_dr call, which is 120 credits or $0.012, and the editors run refdomains on perhaps 150 doubtful domains at 2 credits each, adding 300 credits. Total: 420 credits, about four cents. The $100 starter credit is 1,000,000 credits. Tiny.
The money is not the constraint. Human attention is, and so is the fact that the data refreshes about monthly, which makes a daily schedule wasteful because the numbers will not have moved. The bulk workflow example shows one request from start to finish. Read it.
Scheduling and retries
Keep it boring. One bulk call per draft, run when the draft is saved for review, with a sensible backoff on transient errors and a hard stop on the ones that cannot fix themselves. That is the whole design.
A 401 or 402 should never be retried in a loop, since neither fixes itself, so alert a person instead. For larger audits of existing pages, batch them into groups of up to 2,000 domains, sort by score and hand the bottom of the list to an editor rather than reading everything. Calm and cheap.
Because the index refreshes about monthly, a re-score of the same domains within the same month will usually return the same figures. Cache them. Pair the job with the review checklist so automation supports the human steps and does not replace them.
How this compares with Ahrefs
Fair comparison, based on public pricing and our own limits.
Billing model
Ahrefs sells API units inside subscriptions, from $129 a month for 200,000 units, with at least 50 units per request. We bill prepaid credit per call with no subscription or minimum.
Where Ahrefs or Semrush are fuller
Keyword research, rank tracking and daily-fresh crawling are theirs. Use them if you need those.
Where this fits
Cheap, bulk link data for sanity checks, run as often as your pipeline needs.
Metric difference
Our DR is a public log formula over referring domains. It is comparable in spirit to Ahrefs DR but a different number.