Semrush vs Ahrefs on AI visibility: an index of 495 million prompts against 25 tracked prompts of your own

Yes, both of them track AI visibility, and that is the least useful true sentence available on this subject. Ahrefs and Semrush have each added AI answer tracking, and the two products answer different questions. Ahrefs sells you an index: its Brand Radar page offers to "See any brand's AI visibility instantly across 493 million prompts" and calls itself "The largest AI visibility database". Semrush sells you a subscription to a set of prompts you write yourself: its AI visibility pricing page puts "Track 25 custom prompts" on the Base plan, priced per domain. One tells you what the market's answers look like for questions somebody else collected. The other watches twenty five questions you chose. Decide which of those you need and the comparison takes ten minutes.

What each document published on 21 October 2026

Both pages were opened on the same morning and quoted from the document the server sent.

The line that decides it	Ahrefs Brand Radar	Semrush AI visibility, Base
Surfaces named on the page	AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini, Copilot	"Mentions from ChatGPT, Google AI, Gemini, and Perplexity"
Claude	Named, in the line "Claude consumes 8 checks per update"	Not named anywhere on the page
What you are buying	An index of 493 million prompts, plus your own Custom Prompts	"Track 25 custom prompts", with daily AI rankings
Unit the price is per	Inside an Ahrefs plan, with Custom Prompts listed separately	Per domain, billed annually, with a monthly toggle
Cadence stated	"Track prompts daily, weekly, monthly" and "Each prompt is checked daily on every selected platform"	"Daily, weekly, and monthly data updates"
Seats	Not stated on the Brand Radar page	Additional Users is a paid add-on
Single sign-on	Not stated on the Brand Radar page	Enterprise only: "SSO, team governance & audit logs"
API	Not stated on the Brand Radar page	Enterprise only: "Custom integrations & API"
White label	Not stated on the Brand Radar page	Through the Pro Report add-on, not on Base

One thing on the Ahrefs page is worth pinning before you quote it to anybody. It marks Custom Prompts as "Free for every paid Ahrefs Lite plan and above" and also lists a starting monthly price for the same feature. Both are in the document the server sent. Ask which applies to the plan you hold, in writing, and keep the reply.

The arithmetic that decides which of the two you can use

Write your own number down before you open either page. Count the questions a buyer of your product would actually type, with your company name absent from every one of them. Multiply by the surfaces you want read. Multiply by how often you want them re-read.

Thirty questions on five surfaces, read weekly, is 600 recorded answers a month. That number, not the plan name, is what you are shopping for. Against it, "Track 25 custom prompts" is a hard ceiling you will reach in the first week if you sell more than one thing, and an index of 493 million prompts is not a substitute for it, because an index holds the questions somebody else collected rather than the thirty you care about. Two honest products, two different jobs.

This is also where the suite question answers itself. A suite add-on is the right purchase when the AI reading is a column you want next to your keyword and backlink data, and when a small tracked set covers your category. A separate product earns its own invoice when the tracked set is the work: many questions, several markets, every surface, re-read often enough that you can tell a change from a wobble.

The per-engine weight, which is where the real meter shows

Every vendor in this category meters the expensive models differently, and only some publish how. Ahrefs publishes it on the Brand Radar page: "Claude consumes 8 checks per update", against one check for its other platforms. Rankscale publishes it in a credit calculator, where Claude is weighted at 2 and ChatGPT, Perplexity, Google AI Mode and Gemini sit at 0.25, which is the same eight to one. Semrush's AI visibility pricing page does not name Claude at all, so on that document the weight question has no answer because the surface is not offered.

We publish ours on our own pricing page: "One credit is one buyer question, asked on one AI app, once", and "A Claude check uses 3, because Claude costs us more to run". Three times our base unit against eight times theirs, and these are not the same meter, so compare the ratio rather than the number. The reason to care is practical: Claude is where a researching buyer often lands, and a tool that prices that surface at eight times its base unit is a tool you will quietly stop asking.

Where AI Knows Us sits, and the four lines where both of these beat us

We are the third option on this page, and the narrow case for us is stated as a criterion rather than asserted. We publish per-engine pricing, we name all five surfaces we read, we ask every tracking question without naming your business, and we write and publish the pages afterwards, which neither of these two suites offers. On our pricing page read the same day, Visibility API is published on the GEO Scale and Everything cards, and the comparison line "Your numbers fed into your own tools" reads Yes on Scale, Agency and Everything.

Now the lines where they are ahead, because a page that left these out would be worth nothing to you.

• Cadence. Ahrefs states each prompt is checked daily on every selected platform. Semrush states daily, weekly and monthly updates. Our own page states "Every plan checks weekly, and you can run extra checks any time", with 30 and 90 days on the questions you have not prioritised. Daily beats weekly and we publish weekly.
• Seats. Semrush sells additional users as an add-on, so the number can grow. Our page reads "Logins included: 1, full access" on all eight plans. One login is one login.
• Single sign-on. Semrush publishes SSO on Enterprise. We publish no SSO anywhere on our pricing page, and a procurement team that requires it will stop reading at that line.
• An index. Ahrefs holds 493 million prompts and we hold none. If the question you want answered is what the answers in your category look like for brands other than yours, buy the index.

If the Semrush module is the one you are weighing, the tier where its published price sits is the thing to settle first, and that tier question has its own page. It carries no durable inbound link from anywhere on this site, which is true of 963 of our 1,987 live pages and of 175 in this category, counted across 2,075 stored bodies.

Questions buyers ask about Semrush, Ahrefs and AI answer tracking

Does Semrush track AI visibility?

It does. Its AI visibility pricing page publishes a Base plan described as being "For businesses that want to track how they're mentioned in AI search", with mentions from ChatGPT, Google AI, Gemini and Perplexity, 25 custom prompts tracked, and daily, weekly and monthly data updates. The price figure is published in the page, per domain, billed annually.

Does Ahrefs track AI visibility?

It does, through Brand Radar, and the design is index first: the page offers visibility for any brand across 493 million prompts, with your own Custom Prompts checked daily on each selected platform. The surfaces it names are AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini and Copilot.

Is the module inside my SEO suite enough?

It is enough when your tracked set is small, your market is one country, and you want the AI column beside your keyword data. It stops being enough at the point where the tracked set is the deliverable: thirty or more questions, several cities or countries, every surface including the metered ones, and a record you can hand to somebody else. Work out your own question count first, because that single number settles it faster than any feature table.

Which one do I buy if all I want to know is whether I get named?

Buy on the tracked prompt allowance and on whether the raw answers come out as a file. Appearing or not appearing is a yes or no per question per surface, so what you need is enough questions to make the fraction mean something, and the answer text itself so you can see which competitor took your place. A visibility score with no answer text behind it tells you that something moved without telling you what to publish.

Why do the two suites report different numbers for the same brand?

Because they asked different questions. An index reports on prompts collected by the vendor, and a tracked set reports on the prompts you wrote. Two different question banks produce two different fractions for the same company on the same day, and both are accurate about what they measured. When you compare, compare the question bank first and the number second.

Does either of them write the pages that close the gap?

Neither page offers that. Both are measurement products, and the publishing work stays with you or with an agency. That is the line we sell on, and it is also the line to check hardest in any demo: ask who writes the page, who approves it and who puts it live.

The ten minute check that settles this before any demo

Open a blank document. Write the questions your buyers type, with your name in none of them, and count them. Write the surfaces you want read, and mark the ones you would be annoyed to lose. Write how often you want a fresh reading. You now hold three numbers and the choice resolves against them: a small set beside your SEO data points at the suite module you already pay for, and a large set across several markets with metered surfaces points at a product that does nothing else. Take those three numbers into the demo and ask each vendor to price them, in writing, with the per-engine weight named. The vendor that answers that in one email is telling you how it will answer everything else.