Mistakes and fixes
Competitor backlink analysis through an assistant: mistakes and fixes
Gap analysis is where assistants go wrong fastest, because the question sounds simple and the data is large. Seven mistakes we see, each with the repair.
SEO AI Pal research team ·
What you can really do for free
A competitor backlink analysis tool free of charge exists in two layers here. The domain, overview, stats and suggest actions need no key, so an assistant can size up several rivals at zero credits.
Paid depth starts small. compare takes up to 20 domains for 1 credit, intersect takes 2 to 10 domains for 2 credits, and each call costs the same however many rows return. One credit is $0.0001.
Do the ranking first and the digging second. Compare ten rivals, keep the three nearest your own rating, then intersect those.
Rivals worth checking are rarely the famous ones. They are the three or four sites that already sit next to you in the results for the queries you care about, and a short list keeps both the credits and the reading time small.
Seven mistakes, each with a repair
Every one of these has cost someone an afternoon.
1. Naming the competitor in words
"The big blue shoe store" is not a domain. Make the assistant confirm each domain before any call.
2. Comparing a brand to a subdomain
Mixed levels give meaningless ratings. Send bare registrable domains only.
3. Trusting a summary of 5,000 rows
Long output makes models skim. Ask for the ten strongest referring domains, then page on purpose.
4. Treating overlap as a to-do list
A site linking to two rivals is a lead, not a promise. Read a few pages before outreach.
5. Forgetting the refresh
A rival's link from last week may be absent. Say the data is monthly.
6. Calling the rating a ranking factor
Our DR is a log of referring domains. Google never used it. Use it to sort, not to predict.
7. Letting the model decide what is toxic
Models label links as spam on vibes. Google's SpamBrain (announced December 2022) works at scale on signals you cannot see.
A bulk call you can run before you ask the chat
One credit buys the ratings of up to 2,000 domains with bulk_dr. For compare and intersect, send a two-domain test first and read the reply before you build a prompt around it.
# 1 credit: bulk DR for a short list
curl -H "Authorization: Bearer sk_live_..." \
"https://seoaipal.com/api.php?action=bulk_dr&domains=example.com,example.org"
# free first look at a single rival
curl "https://seoaipal.com/api.php?action=overview&q=example.com"Worked example (hypothetical)
Three rivals, forty credits, one afternoon
Say you sell hiking boots. You compare ten rivals for 1 credit, keep three, and intersect them for 2 credits. Add six refdomains and anchors calls at 2 credits each. The total sits around 15 credits, about $0.0015. These are example figures, not a customer story.
- Compare first, so intersect only runs on the shortlist.
- Ask the assistant to quote
_creditsafter every call. - Read ten linking pages by hand before you email anyone.
Where this stops being enough
Link data answers who links to whom. It does not say what ranks. Keyword gaps, traffic estimates and daily-fresh crawls are jobs for Ahrefs and Semrush, which are the fuller tools there.
Pair them honestly. Keyword research there, link lookups here, and a prompt from the runbook to keep the chat on the receipts. The spend is explained in the budget page.
Before you email a single site from the overlap list
- Open five of the linking pages in a browser and read them.
- Check that the page is live and indexable, not a thin directory.
- Look at the anchor text pattern with the
anchorsaction, 2 credits. - Note the date. Our index refreshes monthly, so a link may already be gone.
- Decide what you would offer that site, in one sentence.
- Skip anything you would be embarrassed to explain to a client.