Is a check backlink tool the same as a rank tracker?

No. It reads links: referring domains, anchors, top pages and a 0-100 rating. Keyword positions and daily crawls belong to Ahrefs or Semrush, which are fuller products for that job. Use each where it is strong.

How many rows can one call return?

Up to 1,000 rows per call, with page and per (10 to 5000) for paging, and deep pagination capped at 50,000. The call is billed once whatever the row count, so a larger per does not cost more.

Why does my DR differ from Ahrefs?

Our rating is round(100 log(ref_domains+1) / log(210^7)), a log scale on referring domains. Ahrefs DR is a separate metric with its own method. Similar in spirit, never the same number.

Troubleshooting

Check backlink tool calls failing in your assistant: a symptom table

Assistants smooth over errors. This page lists what you see in the chat, what happened underneath, and the fix, starting with the symptoms that cost the most.

SEO AI Pal research team · Updated 2026-10-29

Symptom, cause, fix

Statuses are the ones the API returns: 401 for a missing or bad key, 402 when credit runs out.
What the chat shows	What happened	Fix
"I could not retrieve data" with no detail	Probably HTTP 401. The key is missing, mistyped or pasted with a trailing space.	Re-paste the key. Test with the free `overview` action in a terminal.
Works yesterday, "service unavailable" today	Likely HTTP 402. Balance reached zero after a loop or a long session.	Top up with USDT (TRC-20) from the account page. Free actions still answer.
A rating of 0 or no rows for a real site	The index may not hold the domain yet, or the name was sent with a path or `https://`.	Send the bare domain. Retry next month if the site is brand new.
The new link you built is missing	The index refreshes about once a month.	Wait for the refresh. Do not re-run the call ten times.
Numbers differ from another vendor's dashboard	Different crawlers and different scoring. Our DR is its own metric.	Compare trends, not single figures.
The reply lists far fewer rows than expected	The schema capped `per`, or the assistant read only page one.	Ask for `page=2` explicitly.
The reply has numbers but no `_credits`	No call was made. The model answered from memory.	Apply the rules in the prompt runbook.
Charges higher than you planned	The assistant retried a paid action, or looped over a list.	Cap retries to one and batch with `bulk_dr`.

Why the status code matters more than the sentence

An assistant turns a 402 into a polite paragraph. You read the paragraph, not the code. Ask for the raw status. It is the cheapest diagnosis you have.

Split the fault in two. Before you blame the model, run the same URL in a terminal. If curl returns JSON with ok true, the API is fine and the fault sits in the wrapper or the prompt. If curl fails too, the key or the balance is the problem, and no prompt wording will fix it.

Ninety seconds of triage

Run the failing call with curl and a free action. Does ok come back?
Run it with a paid action. A 401 points at the key, a 402 at the balance.
Print the domain exactly as the assistant sent it. Look for www., paths and quotes.
Check the date of your last order against the monthly refresh.
Read the tool-call log of your assistant product, not its summary.

Mistakes and fixes in the wrapper itself

The same three mistakes show up in most wrappers. Putting the key in the URL means it appears in logs and in shared transcripts, so use the bearer header where your product allows it.

Exposing every action is the second. Your GPT only needs four of them, and each extra action is another way to spend credit you did not plan on. The third is requesting 5,000 rows when a reply uses ten.

None of these are exotic. Fix them once and the table above mostly stops applying. Wrong numbers that survive all of it are covered in the myths page.

Quick answers

Is a check backlink tool the same as a rank tracker?: No. It reads links: referring domains, anchors, top pages and a 0-100 rating. Keyword positions and daily crawls belong to Ahrefs or Semrush, which are fuller products for that job. Use each where it is strong.
How many rows can one call return?: Up to 1,000 rows per call, with page and per (10 to 5000) for paging, and deep pagination capped at 50,000. The call is billed once whatever the row count, so a larger per does not cost more.
Why does my DR differ from Ahrefs?: Our rating is round(100 log(ref_domains+1) / log(210^7)), a log scale on referring domains. Ahrefs DR is a separate metric with its own method. Similar in spirit, never the same number.