Using Ahrefs MCP With AI Every Day: How I Save Up to 83% of the Units per Call

Background: AI Plus Ahrefs MCP Is My Daily SEO Run Now

These days almost all of my SEO daily run starts from an AI chat. I ask for the SERP of a keyword, the volume of a keyword list, whether a page ranks for a term, or the top pages of a competitor, and the AI calls Ahrefs through the MCP connection and gives me the answer in the same window. No more switching tabs, exporting CSVs and copy pasting into a spreadsheet.

Honestly, it is really helpful. Things that used to take me 20 minutes now take one prompt. And because it is so easy, I started calling Ahrefs for everything, haha.

The Problem: Every Call Costs Units

Then I looked at the usage page.

The Ahrefs API does not bill you per request. It bills you per row that comes back, and every column you select has its own price per row. Base columns like the keyword, the position or the URL are cheap. Columns like search volume, keyword difficulty, traffic potential or traffic cost around 10 units per row each. On top of that, a call has a minimum charge of 50 units, even when it returns almost nothing.

So one innocent prompt like “check the top 10 for this keyword with DR and traffic” quietly becomes 20 rows multiplied by several priced columns. For a SERP check I did for a blog draft, that was 300 units for one keyword. Multiply that by a full content batch and the monthly cap is gone before the month is half over.

So the question for me was simple: can we save it?

Working in an AI First Company Means a Lot of Trial and Testing

I work in an AI first company, so my days are full of trying things with AI, breaking things, and testing again. Ahrefs units became one of those experiments. I started logging every call in a simple ledger (date, project, endpoint, rows returned, units spent), and after a few weeks the pattern was clear. Most of the units were not spent on the data I used. They were spent on data I never looked at.

From that trial and testing, two habits stayed.

#1 Solution: Ask Only for What Matters, Exclude the Rest

Every request to Ahrefs can be divided into the things I really need for the decision and the things that are “nice to see”. The second group is where the units leak. So before every call I decide which columns and rows I will actually use, and I exclude everything else.

These are the things I exclude from the first call, every time:

• Domain Rating and traffic columns in a SERP check. For a blog draft I only need to know who ranks and what their page is about. Position, URL, title and result type are enough.
• Top 10 results. I take the top 5 and analyse the first 3 organic results. Positions 6 to 10 never changed my brief.
• AI Overview, People also ask and other SERP features as rows. They are billed as rows too. Filtering the result type to organic removes them.
• Monthly volume history. It is priced per historical month with its own minimum, so I never ask for it unless the trend is the actual question.
• Keyword difficulty and traffic potential when I am not going to report them. Every priced column I drop saves 10 units on every row.
• One call per keyword. If I need metrics for 20 keywords, they go into one batched call, not 20 calls.

One small trap I found along the way: if you ask for only the top 3 positions, you might get only one organic result back, because AI Overview and snippets also take position numbers. Asking for the top 5 and filtering to organic gives me the 3 real competitors I need.

Here is the same SERP check, before and after:

#2 Solution: Always Save the Output From Ahrefs

The second habit is boring but it matters a lot. Every response from Ahrefs gets saved, raw, every row, before any filtering. It goes into a local cache folder, one file per call, named by project, endpoint and keyword, and one line goes into the ledger.

Why? Because the AI can read that cache. The next time I ask for the same keyword, or a similar one for another article, the AI checks the cache first and reuses the data instead of paying for it again. The cheapest Ahrefs call is the one you do not make.

I learned the “every row” part the hard way. On one project, we pulled two long keyword lists (more than 300 rows together), kept only a shortlist of 58 keywords, and threw the rest away. A few weeks later we needed the long tail, and the only way to get it back was to pay for the same pull again. Now the rule is clear: the shortlist is an extra file next to the raw response, never a replacement for it.

Impact: Up to 83% Fewer Units per Call

The SERP check for a blog draft went from 300 units to 50 units. That is 83% fewer units for the call I make most often.

To see it at a bigger scale: in one content batch we prepared 81 articles, one lean SERP call per article. The whole batch cost 4,050 units. With the old shape of around 300 units per keyword, the same batch would have cost around 24,300 units. And when an article topic was already pulled before, the cache answered it for free.

To be fair, 83% is the number for the SERP check, which is my most frequent call. Other calls save less or more depending on how many priced columns you drop. But the principle is the same for all of them: fewer rows, fewer priced columns, never the same pull twice.

What I Learned From My Case

First, AI makes calling an API so easy that you stop seeing the cost. The convenience is the risk. If you use Ahrefs MCP every day, log your spend for a few weeks and you will probably see the same pattern I saw.

Second, the savings did not come from a clever trick. They came from deciding before the call what I will actually use, and from never throwing away data I already paid for.

Get the Skill I Use

I turned both habits into a skill for Claude, so the AI runs the checklist by itself before every Ahrefs call: a free budget check first, cache before pull, the cheapest endpoint, minimal rows and columns, and saving every raw response with a ledger line. You can download it for free here: Ahrefs API Efficiency skill on GitHub.