You have a competitor’s store open in one tab and an empty spreadsheet in the other, and the list is due Friday. On 25 August 2026 we ran that exact job for a Shopify apparel store. One sweep consumed 7,969 of an 8,000 unit Ahrefs API budget in an afternoon, and the budget reset was nine days away.
Keyword extraction means pulling the search terms a site or a market already ranks for into a file you can sort. One row per term, with volume, difficulty and the page holding the position.
So the tool question is really a row question. A row cap is the most keyword rows a plan will show you, or let you download, in one report. How many rows you see, how many you can take out, and what the second thousand costs.
For keyword extraction, Ahrefs wins on depth and Semrush wins on tidiness. Ahrefs publishes the bigger index, a click count per keyword, and a row cap for every plan, so you can size the job before you pay. Semrush builds the groups for you and exports each one to its own tab, which removes a manual sorting step.
| Criterion | Ahrefs | Semrush |
|---|---|---|
| Keyword index, vendor's own figure | 28.7 billion keywords filtered from 110 billion discovered, across 217 locations | over 27.3 billion keywords |
| Intent data | four intents plus Branded and Local, available only on keywords that have SERP data | four intents in the Intent column of the Keyword Magic Tool |
| Clustering | Parent Topic points a keyword at the broader topic that outranks it | up to three levels of subgroups, exported as separate tabs in one Excel file |
| Traffic Potential | organic traffic the number one page earns from every keyword it ranks for | not named in the Keyword Magic Tool documentation |
| Export caps | 250 rows per report on Starter with exports switched off, then 2,500, 30,000, 75,000 and 150,000 rows | free plan shows 10 results and exports 0; no paid-plan row cap published |
| Entry price | $29 a month Starter, $129 a month Lite | $117.33 a month on the SEO plan, billed annually |
Both figure sets come from the vendors, not from us, and vendor figures move. We check them before we quote them, the same way we check the numbers in SEO and AI search work before anything goes to a client. If you want the checking procedure itself, we publish how we review SEO content before publication.
This is for the operator running one or two stores. You need a keyword set out of a competitor and into a sheet, without paying for a second subscription to finish the job. It is not for an agency running 40 client stores. Seat pricing, not row caps, decides that case: extra users cost $40 a month on Ahrefs Lite and start at $45 a month on Semrush.
Ahrefs tells you the row cap before you pay
The cap is the number that matters, because keyword extraction is an export job. Ahrefs publishes the cap for every plan on its pricing page, read on 21 September 2026. Lite gives 2,500 rows per report at $129 a month. Standard gives 30,000 at $249, Advanced 75,000 at $449, Enterprise 150,000. Export rows run from 500,000 a month on Lite to 4 million on Advanced.
The $29 Starter plan is the trap. Ahrefs’ own Starter plan article states 250 maximum rows per report, 200 credits a month, one month of historical data, and export rows per month: none. You can look at keywords on Starter. You cannot take them anywhere.
Depth is the other half. Ahrefs shows a Clicks figure next to volume, built from clickstream data, which is browsing data bought from panels of real users. Its own worked example is the keyword “seo”: 125,000 searches a month, 59,000 clicks. Half that demand never leaves the results page. For a store deciding whether a collection page is worth building, that gap is the decision.
Ownership is simple and slightly annoying. Every paid plan includes one user. More seats cost $40, $60 or $80 a month each, by plan.
Semrush does the grouping before you export
Semrush’s Keyword Magic Tool holds over 27.3 billion keywords by its own documentation. Its advantage is what happens between the search and the file. It sorts results into up to three levels of subgroups by topic, and it tags rows with one of four intents. You can export with the groups kept, and each group lands on its own tab in the Excel file.
That is a genuine hour saved. If you extract 3,000 terms out of a competitor and then have to cluster them by hand, the clustering is the work, not the pull.
The free plan is a demo, not a tier. It shows 10 results in the interface, exports 0, and allows 10 reports a day. Paid plans start at $117.33 a month on the SEO plan billed annually, with 5 sites and 500 keywords tracked daily. The Advanced plan is $455.67 a month for 40 sites and 5,000 keywords.
Here is the gap you should price in. We checked the Semrush pricing page, the subscriptions article and four knowledge base pages on 21 September 2026 and found no published rows-per-report or rows-per-export figure for the paid plans. It may exist in the product. It is not in the documentation a buyer reads first.
If you pull by API rather than by hand, the two price models are not comparable. A unit is the metered credit each vendor bills API calls against. An Ahrefs unit and a Semrush unit are not worth the same thing, so never compare the totals. Ahrefs’ consumption docs charge the larger of a 50 unit base or the fields you select multiplied by the rows returned. Semrush’s keyword reports charge a flat unit price per line, fixed per report type.
| What you are pulling | Ahrefs API v3 | Semrush API v3 |
|---|---|---|
| How a request is charged | the larger of 50 units, or fields per row multiplied by rows | a flat unit price per line, set per report |
| A keyword overview | 1 unit per field per row, 5 or 10 for expensive fields | 10 units per line |
| Related or broad-match terms | priced the same way, by field | 40 units per line related, 20 broad match |
| Keyword difficulty | priced as one more field | 50 units per line |
| Running the same pull twice | cached requests consume nothing | not published |
| Ceiling on one report | your plan's row cap | 100,000 results, then paginate |
We run Ahrefs, and we can tell you what a real pull costs
Autonomous Technologies builds and runs the systems behind Shopify and Shopify Plus stores. That means the order, stock and reporting plumbing between your storefront and your back office. Search data is one input to that work, so we buy keyword tools the way you do.
We do not run Semrush. Every keyword dataset in our workspace is an Ahrefs API v3 pull. So read the Semrush column above as documentation we verified, and the Ahrefs column as documentation we also live with.
| The pull | Date | What it cost | What it showed |
|---|---|---|---|
| Keyword sweep for a Shopify apparel store | 25 Aug 2026 | 7,969 of an 8,000 unit budget | a colour cluster worth about 6,060 US searches a month with no page to land on |
| Defence parts market feasibility | 20 Sep 2026 | 8,333 units over 24 pulls, measured at 32 units per row for a five-field overview | 40 of 114 seed terms, the phrases we started from, returned no row at all |
| Karachi home services market | 4 Apr 2026 | not recorded | seven head terms totalling about 1,130 searches a month, with the local-language tail below the reporting threshold |
| Check of our own site | 17 Sep 2026 | not recorded | 5 of 7 keyword volumes quoted in an internal analysis were more than 30% out against the pull |
The Karachi pull became the SafaiKaro local search engine, where pages earning search impressions went from 24 to 40 across 35 days either side of the 23 July 2026 release. The keyword extraction was one afternoon of that project. Deciding which of those terms deserved a page took a week.
That is the honest shape of the choice. The tool bill is small next to the hours after the export, which is why the row cap and the clustering matter more than the sticker price.
Never quote a volume you have not pulled yourself. On 17 September 2026 we checked seven keyword volumes used in a strategy document against a fresh pull. Five were out by more than 30%, in both directions. Both tools estimate, and estimates drift between refreshes.
Where this breaks: the tail your store actually sells into
Both tools report what they can measure, and neither reports the tail. In the defence pull on 20 September 2026, 40 of 114 seed terms came back with no row. That included every phrase describing the product itself. On the Karachi market in April 2026, the local-language terms customers actually type sat below the reporting threshold of roughly 10 searches a month. They were invisible in both tools.
For a Shopify store that means product-specific and regional language will under-report or vanish. Neither tool is lying. A term with 30 real searches a month from buyers can be worth more than one with 3,000 and no buying intent. Only one of those two shows up in an export. The route in is covered in mining profits from zero search volume keywords. Take the term from your own site search and support tickets first. Then use the tool to confirm the SERP, not the demand.
One more break, on the ownership side. Both tools price a seat, not a workflow. If three people need the export, extra seats cost $40 to $80 a month each on Ahrefs, or from $45 a month on Semrush. That is before anyone opens the file.
Questions about Semrush, Ahrefs and keyword extraction
Is Semrush or Ahrefs better for keyword research?
Semrush suits research that feeds content planning and paid together, because intent labels and topic subgroups arrive in the same view and survive the export. Ahrefs suits research scoped to organic only, where Traffic Potential and the Clicks count matter more than raw volume. If you are choosing one and the output is a spreadsheet, the published row cap is the tiebreaker.
What is Ahrefs' return rate metric?
Return Rate is Ahrefs’ measure of how often the same people repeat a search within a 30-day window. In Ahrefs’ own words it is “a relative number”, not a count. A value of 1 means people never search that keyword again. A value of 2 does not mean twice a month, only that they search it quite often. Read it as a habit signal: a high return rate points to a navigational or news-driven query, a low one to a question answered once and never asked again.
Which tool has the bigger keyword database?
Ahrefs, on the published figures we read on 21 September 2026. Ahrefs states 28.7 billion keywords filtered from 110 billion discovered, across 217 locations. Semrush states over 27.3 billion. The gap is about 5%, which is far smaller than the gap between either database and your own site search log.
Can you export keywords from both tools?
Yes, on paid plans, with different ceilings. Ahrefs publishes 250 rows per report on the $29 Starter plan, with exports disabled entirely. Then 2,500 on Lite, 30,000 on Standard, 75,000 on Advanced and 150,000 on Enterprise. Semrush exports to XLSX or CSV with groups preserved as tabs. It publishes no paid-plan row cap that we could find, only the free-plan limits of 10 results shown and 0 exported.
Do I need both Semrush and Ahrefs?
No, for almost every single-store operator. Two subscriptions plus two sets of seats buys you a 5% database difference and two exports to reconcile. The one case that earns both is a team already buying Semrush for paid search and reporting, that then needs Ahrefs’ click data and Traffic Potential for organic decisions. That is a workflow split, not a data gap.
Buy the row cap you need, then spend the saved hours on the sheet
Decide the job first. If the deliverable is a grouped content plan, Semrush hands you the grouping. If the deliverable is a large sorted file and a decision about which pages to build, Ahrefs tells you what you can take out before you pay.
Count the rows you actually used before you renew either one
Here is the real hesitation, and it is fair. You already pay for one of these, and nobody wants a second subscription or a third opinion on a $129 line item. So do not buy anything yet. Before either subscription renews, open last month’s export and count the rows you actually used. If the number is under 2,500, the cap is not your problem and the sorting afterwards is.
The tool is the cheap part. The expensive part is what happens after the export: who owns the sheet, which figures get quoted, and whether anyone re-pulls them before they reach a plan.
