An Ahrefs alternative, told as a diary: where 6,600 small lookups a month went

Open ahrefs.com this autumn and the first thing on the screen is not Ahrefs. A cream-coloured band introduces Letaido, another product from the same company, with a Get Started button. Below it, on Ahrefs blue, the navigation offers Product, Our data, Resources, Pricing and Enterprise, and the headline talks about search and AI. The word “backlink” is nowhere on that first screen. Ahrefs calls itself a marketing platform now, and its API comes as one feature of a subscription, not as a product with its own price list.

That matters to the people in this diary, who do not exist and have never paid Ahrefs a cent. Picture three colleagues at a small link-building agency: a developer, an outreach lead and an account manager. They are on Ahrefs Standard, and the developer has wired API v3 into an internal page everyone calls the prospect card. Paste a domain, and it shows who links to that site and with which anchors before anyone writes a pitch email.

The team is invented; what it runs into is not. Plan prices, unit allowances, row caps and cost rules are Ahrefs' own, read from its pricing guide, help centre and developer docs on 8 October 2026. Every figure about ahrefs.com's links is what the apibacklinks index returned that day.

The account manager asked this morning why our Ahrefs API units keep running short before the month does. So I did the sum I should have done in spring.

Ahrefs meters API v3 in units. A request costs whichever is larger: a base of 50 units, or the per-row cost of the fields you select multiplied by the rows returned. Most fields cost 1 unit; some cost 5 or 10. Our two jobs sit at opposite ends of that rule.

The 12 is not my number. It is the per-row cost in the worked example in Ahrefs' docs, where a 500-row backlinks request comes to 6,000 units, and our column list was about that greedy. The 100 is a floor: a card that asks two small questions pays the 50-unit minimum twice, even when ten rows come back. Standard includes 800,000 units a month. On paper we were 340,000 over, every month.

Leaving Ahrefs was not the first idea. The first idea was to be a better Ahrefs customer.

• Trim the columns. Cost follows the fields you ask for. The all-backlinks endpoint has 82 selectable fields; the two traffic fields cost 10 units each and four others cost 5. Cutting the weekly report down to one-unit columns shrinks its bill a lot. It does nothing for the card, which already sits at the minimum.
• Count on the cache. Requests that Ahrefs serves from cache consume no units. Pleasant when two colleagues look up the same domain, impossible to budget around.
• Use the free endpoint. Ahrefs keeps a free Domain Rating endpoint. Since 10 August 2026 it wants an API key, but its calls still consume no units. Fine for a score; the card needs lists.
• Move up a plan. Advanced is $449 a month against $249 for Standard and carries 2,000,000 units, with 500 rows per request instead of 250. Extra units can also be charged as you go on plans that support it; the price of one is not published in the pricing copy we read.

Nothing in the Ahrefs rule is unfair. The meter suits the request that returns many rows, where a 50-unit minimum disappears into the total. Our card is the opposite animal: thousands of tiny requests, each one set off by a person pasting a domain. The shape of our traffic was wrong for the meter, not the meter wrong.

Every prototype we ever wrote against Ahrefs used ahrefs.com as its target, for a thrifty reason: on a paid plan, test queries against ahrefs.com, yep.com or firehose.com consume no units, with the limit capped at 100 rows. So when I opened apibacklinks, a backlink API built on Common Crawl data, my fingers typed the same domain.

The overview action needs no key. It came back with a DR of 67, 83,108 referring domains and 1,690,141 backlinks, 79.2% of them dofollow, plus the top five referring domains, anchors and pages. Then I spent two credits on refdomains and read the list the way our outreach lead reads a prospect.

It is a strange and instructive list. The largest referrer by link count is ppcrao.in, with 75,139 links and a DR of 19. Fourth is ednc.org: 25,728 links from 10,019 pages, all carrying one sentence-long anchor that invites readers to learn about canonical tags, all pointing at a single Ahrefs blog post. One sitewide helper link outweighs most editorial coverage. HubSpot, by contrast, sends 2,453 links, 2,330 of them dofollow. That is exactly what the card exists to show: how many links, from how many pages, followed or not. Anyone can browse the same report without signing in on the explorer page for ahrefs.com.

The port took an afternoon, mostly because the two APIs agree on the boring parts. Ahrefs takes GET requests at https://api.ahrefs.com/v3/{tool}/{endpoint}, with the key in an Authorization: Bearer header, and answers in JSON. apibacklinks takes GET requests at one endpoint, /api.php?action=…&q=domain, accepts the same Bearer header or a key parameter, and answers in JSON. Our HTTP wrapper changed a base URL and a secret.

The real difference is in how you ask. Ahrefs has a select parameter that names the columns you want and, through them, sets the price; filters go in a JSON where expression. apibacklinks prices the call instead of the columns, and its backlinks action filters by follow or nofollow, link type, source TLD, anchor text, minimum donor DR and date range. A list call costs 2 credits whether it returns ten rows or a thousand. The API reference lists all 19 actions. I deleted our Ahrefs column-budget helper with some ceremony.

Two things broke after the card stopped calling Ahrefs, and neither was code.

The first was the outreach lead's rule of thumb, a minimum Ahrefs Domain Rating for any site worth a pitch. The DR on the new card is not that number. apibacklinks computes its own 0–100 logarithmic score from referring-domain counts in its own graph, and on that scale ahrefs.com itself is a 67. You cannot turn one into the other by adding a constant. The profile I had just pulled shows where the scale sits: of the top 100 domains linking to ahrefs.com, 42 fall between DR 15 and 29, 28 are below 15 and only 3 reach 70. So we rebuilt the threshold from evidence, running last quarter's accepted and rejected prospects through bulk_dr, and relabelled the field so nobody would quote it to a client as an Ahrefs figure.

The second was a guest post. The account manager looked up a placement that had gone live two weeks earlier and found nothing. That is the index working as described: it is rebuilt as Common Crawl releases arrive, roughly monthly, and a new link can take weeks to appear. The newest sighting of any link to ahrefs.com in it was dated 9 September. Ahrefs says its index of live backlinks is updated every 15 to 30 minutes. For the question of whether a promised link has gone up, those are different instruments.

By the end of the month the split had drawn itself, and most of our work stayed with Ahrefs.

• Anything about this week. The weekly new-links report and every is-it-live-yet check need a crawler that came by recently. Ahrefs runs its own, AhrefsBot, and publishes figures to match: 285.7 million domains after vetting, 493.9 billion pages, 35 trillion external backlinks in its history, 16 years deep. The apibacklinks index holds 181,132,097 domains from a broad public sample of the web. It is smaller and less fresh, and says so.
• Anything about the past. The Ahrefs history endpoints for referring domains and Domain Rating have no counterpart in the cheaper API.
• Anything that is not a link. Keyword research, rank tracking and site audits come with the same Ahrefs subscription, and Brand Radar's data on visibility in LLMs sits in the same API.
• The interface. Two of the three of us never touch an API. They work in Site Explorer all day.
• The name. When a client's contract says Ahrefs DR, only Ahrefs can supply it.

Some teams should not move a single call. If your usage is a handful of large pulls a month, the Ahrefs unit model treats you well, because the minimum charge vanishes inside a big request. If your calls already fit inside the units your plan includes, the API costs you nothing extra and migration is pure work. Our problem was narrower than the price of Ahrefs. It was one chatty job.

The card now makes two list calls to apibacklinks, refdomains and anchors, at 2 credits each. A credit is $0.0001, so a card costs $0.0004 and 6,600 cards come to 26,400 credits: $2.64 a month. Signing in with Google put 1,000,000 credits on the account, which this job would take about three years to spend. There is no subscription behind it and no monthly minimum.

On the Ahrefs side the weekly report still burns its 480,000 units, now with 320,000 to spare inside the Standard allowance. The move to Advanced, $200 a month more, is off the agenda.

What we gave up belongs in the log too. The card draws on a smaller sample of the web than it did in August. Its DR is a different scale, labelled as such. A link built last week will not be on it. For deciding whether a site deserves a pitch email, none of that has mattered yet; for proving a placement to a client, all of it would, which is why that job never left Ahrefs.

We did not leave Ahrefs. We moved 6,600 small questions a month to a place that prices small questions as small. The backlink APIs hub collects the same exercise for other vendors, including DataForSEO.