Ahrefs AI in 2026: Brand Radar, MCP, Ask Ahrefs and What They Cost

Two kinds of AI inside Ahrefs

Ahrefs now sells itself as a marketing platform built to keep your brand discoverable across search, AI and the web, and its AI features, covered here from Ahrefs' own pages as of October 12, 2026, fall into two groups. The first measures how brands appear in AI answers, and Brand Radar is the headline product: it tracks brand mentions and citations across AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini and Copilot. The second group is AI that helps you work with Ahrefs data, from the MCP server and the API to assistants inside the app. Custom prompts come with every paid plan from Lite, and the full AI index costs from US$199 a month ^[1].

How Brand Radar builds its index

Where the prompts come from, and how they run

Brand Radar runs a large set of questions through AI platforms and stores the answers. You then search those answers for any brand, product or term. There is no setup for the index: you search, and the data is already there. Ahrefs does not invent its prompts. It takes questions from Google's People Also Ask, linked to keywords in its database, and expands them with semantic fanout ^[2]. In plain terms, the questions mirror what people really search, plus the sub-questions an AI would need to answer them.

The prompts then run through the free, publicly available web interfaces of ChatGPT, Gemini, Perplexity, Copilot and the other platforms. That mirrors what a typical free user sees, not what a paid subscriber sees, so results for paid AI tiers may differ. Locations follow the country mix of Ahrefs' keyword database, and coverage is strongest in English. Keep both points in mind when a result looks odd for your market.

How big the index really is

The methodology post lists approximate monthly query volumes: about 14.5 million for ChatGPT, 14.6 million each for Perplexity and Gemini, 13.3 million for Copilot, 282 million for AI Overviews, 24 million for AI Mode and 13 million for Grok. Most of the headline number comes from Google's AI Overviews. The chatbots are a much smaller share. If your buyers live in ChatGPT rather than Google, weigh the chatbot rows more heavily than the total.

The total itself is hard to pin down, because Ahrefs' pages disagree. The Brand Radar page says 476 million monthly prompts, the help center says 405+ million, the custom prompts article mentions 350M+, the 2025 launch post said 190+ million, and the What is Ahrefs article gives both 150 million and 239 million. Our reading is that the index has grown fast and older pages were never updated. That matters. A buyer comparing Ahrefs with another AI visibility tool on prompt counts alone could be comparing a 2025 figure with one from last month without knowing it.

The platforms covered

The AI Visibility Index covers AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini and Copilot. Grok was added in April 2026 at US$199 a month, but the help center now says that, following recent policy changes by Grok, Ahrefs is temporarily unable to gather new data. Claude is tracked for custom prompts only, and each Claude check consumes 8 checks. ChatGPT, Perplexity, Gemini and Copilot are refreshed monthly, with a 90-day reporting window, while AI Overviews and AI Mode update continuously with the keyword database. For daily tracking of your own questions, custom prompts are the better tool.

Reading the numbers without fooling yourself

Mentions, citations and pages found

A mention counts once per AI answer that names the brand, however many times the name appears. A citation counts when a page is cited as a source. Ahrefs also tracks pages "found" but not cited, which the AI read in the background without linking. That last metric shows where your content shapes answers invisibly. AI answers sometimes cite links that do not exist, and Ahrefs keeps those in because they reflect real model output. That is honest, and useful. If AI keeps citing a URL on your site that does not exist, redirect it to the right page.

Impressions, Share of Voice and AI adjusted volume

Impressions estimate exposure by adding up the search volumes behind the prompts where your brand appears. AI Share of Voice is your share of impressions against tracked competitors, and Ahrefs warns that it can change a lot depending on how you set up the entities. Since August 31, 2026, Brand Radar uses AI adjusted volume instead of plain Google volume. Each prompt's Google volume is multiplied by the platform's estimated usage: 0.115 for ChatGPT, 0.011 for Gemini, 0.01 for Perplexity and 0.0033 for Copilot.

What the metrics are not

Numbers fetched before August 31 will not match today's, so compare old reports with care. More broadly, Ahrefs is candid about limits. Its methodology says the metrics "are directional indicators, not exact traffic counts" ^[2], and adds that they are not a substitute for audience measurement or traffic analytics. Treat Brand Radar as a visibility audit. It is not a traffic report. For actual visits from AI tools, look to Web Analytics instead.

Setting up a report you can trust

Getting started is quick. Enter your website and click Analyze, and Brand Radar finds your brand name and likely competitors on its own. You can also add brands manually, with several names and websites grouped as one entity. Save the search as a report to keep its brands and filters, but note that a report mapped to a project cannot be moved to another project later. Since August 2026, Brand Radar can match a brand as an entity rather than a text string, which helps with ambiguous names like Apple or Notion. If your brand shares its name with a common word, set it up as an entity before trusting any mention count.

Sudden jumps are not always real. Ahrefs says spikes can come from changes to its own prompt database, and the tell-tale sign is that every brand rises at the same time, including competitors and brands in other industries. In August 2026 the prompt dataset grew from about 14 million to 30 million. Check the chart notes first. Otherwise you may credit your brand with growth that came from Ahrefs' database.

Custom prompts: tracking the questions your buyers ask

Checks, plans and packages

Custom prompts let you track any question you choose, on chosen platforms, locations and frequencies: monthly, weekly or daily ^[3]. One check is one prompt execution on one LLM in one location. Lite includes 150 checks a month, Standard 300 and Advanced 600, while Starter includes none. Extra checks come in three packages: US$50 for 2,500 a month, US$100 for 7,000 and US$250 for 25,000, with overage at US$0.020, US$0.015 or US$0.010 per check. Since January 2026 these packages can be bought without a full Brand Radar subscription. Ahrefs suggests treating the included checks as a built-in trial.

Generating prompts and reading fanout queries

You do not have to write every prompt. Brand Radar can generate custom prompts from your location, language and focus areas such as competitor comparisons or pricing, and prompts can be grouped with tags. Review every generated prompt. AI suggestions tend to be generic until you add instructions such as local intent only, a buyer stage or competitor names, and the form has a custom instructions field for exactly that.

When ChatGPT or Perplexity answers a prompt, it often runs extra searches behind the scenes. Ahrefs shows these fanout queries for both platforms, and most prompts produce about two. Ahrefs explains that content covering those sub-topics is more likely to get cited in AI answers. They are a practical source of content ideas, and among the most useful things Brand Radar shows.

Beyond chatbots: video, forums, crawlers and other reports

Brand Radar also tracks mentions in YouTube videos, Reddit threads that rank in Google, and TikTok. These indexes are in beta and free for all subscribers to Lite or higher plans, but the help center warns that when an index leaves beta it becomes an additional chargeable add-on. Budget for that. AI data also appears outside Brand Radar. Site Explorer's Overview shows AI Overviews history, and Top Pages shows how many AI responses each page appears in if you have a matching Brand Radar subscription. A Looker connector charts AI mentions, impressions and Share of Voice.

The Cited pages report shows which URLs AI answers cite for your topic. Since May 2026 it adds bot visits from Bot Analytics and AI traffic from Web Analytics for your own pages, so you can see which cited pages actually send visitors. Bot Analytics shows which bots crawl your pages, with an AI bots filter across reports; it needs a Cloudflare integration and is free while in beta. Comparing AI crawler visits with citations shows whether being crawled leads to being cited. Pages that are crawled but never visited are worth reworking.

Connecting assistants: the MCP server and the API

Ahrefs' MCP server connects its API to AI assistants such as ChatGPT, Claude and Copilot, with no code required ^[4]. It is available from Lite, and the MCP page lists setup guides for ChatGPT, Claude, Cursor and Copilot Studio. Local MCP setup is no longer supported; only the remote server works. There is no separate fee, but every call consumes API units from your plan, and "One API call costs a minimum of 50 units" ^[4]. Lite includes 200,000 units a month, Standard 800,000 and Advanced 2 million. At 50 units a call, Lite allows at most 4,000 calls a month, fewer for complex requests.

Ahrefs' own examples include keyword research across several reports, comparing how AI answers branded queries over time, and analyzing the backlink profiles of top-ranking pages. The AI-answer example needs the Brand Radar AI Overviews and AI Mode add-on. Start with the backlink example. It uses only core data. Under the MCP server sits API v3, available from Lite, by direct access, MCP or Ahrefs Connect, all sharing the same unit allowance. Rows per request are 100 on Lite, 250 on Standard and 500 on Advanced. Custom endpoints are Enterprise only, some endpoints cost zero units, and API v2 is fully retired.

Assistants and agents inside Ahrefs

Ask Ahrefs, formerly Agent B, is an assistant inside the app. It reads the page you are on and your workspace, and it can run API queries, navigate tools and apply filters on its own. Its price is unclear. The May 2026 post says it is free for everyone during beta, while the homepage says it is free on Lite and above. Expect the stricter version.

Letaido is a separate product: an AI agent that builds reports, dashboards and automations on Ahrefs data. It starts at US$99 a month with US$50 in AI credits, and of each credit payment, 75% becomes credits and 25% is a service charge. It does not consume Ahrefs API units, but it needs an active Ahrefs subscription. In Site Audit, issue pages now have a Fix with Agent A button that sends the issues to Ahrefs' AI agent. If you prefer your own AI tool, a Copy for AI option exports top pages and issues as Markdown, ready to paste into any assistant. Either way, review every suggested fix before publishing it.

AI in content, keywords and audits

AI Content Helper scores a draft against the topics covered by top-ranking pages, on a 0 to 100 scale, and Ahrefs says it identifies core topics rather than counting keywords. Every plan, including Free, creates one new document a month; Enterprise gets five, and more come with the Content Kit add-on. Brand kits, still in beta, turn up to three example articles into written voice guidelines that AI Content Helper then uses to suggest on-brand edits. Edit the generated guidelines before relying on them.

Keywords Explorer has three AI features: AI Suggestions for new keyword ideas, AI Search Intent to group results by intent, and AI Translations for up to 10,000 keywords. All three need Standard or higher, so on Lite you research keywords without them. In Site Audit, Project Boost Pro adds Ask AI and bulk AI detection, and Max adds Batch AI and Patches. Batch AI writes titles and meta descriptions for many pages at once, Patches can publish those fixes without a developer and roll them back, and since February 2026 Patches also generates image alt text.

Ahrefs can also estimate how much of a page is AI-written. Site Explorer shows an AI content level for any page, Page Inspect includes an AI detector from Lite, and bulk detection in Site Audit is capped at 1,000 pages per crawl on Pro boosts. Use it on competitors' content as much as your own. Outside the paid product, Ahrefs offers free AI writing tools, and the free AI Visibility Checker previews Brand Radar. Our guide to what Ahrefs gives away for free covers those tools in detail.

Your data and AI training

The privacy policy names training of Ahrefs' machine learning models among its purposes, and the Enterprise page says Enterprise customers can opt out of having their data used for training. Read the policy with that in mind if you are on a smaller plan. The Terms also warn that AI output may be incomplete, inaccurate or offensive, and they require human review before use. Check every AI draft before it goes anywhere public, from Agent A fixes to Batch AI titles.

A four-week playbook, and what it costs

Start small and let the included allowances tell you whether the paid index is worth it:

• Week one: verify your site and install Web Analytics to see AI referral traffic.
• Week two: set up a Brand Radar report with your brand and three competitors as entities.
• Week three: add 20 custom prompts that buyers ask.
• Week four: read fanout queries and cited pages, then brief content to fill the gaps.

A Lite plan at US$129 a month covers custom prompts and MCP ^[1]. Adding the full AI Visibility Index brings it to US$328 a month, and Letaido adds US$99 more. Brand Radar can be bought on its own, with an Ahrefs Free account included, but Ahrefs' pricing guide says it can only be purchased monthly, except on Enterprise. That means no annual discount for most buyers, so factor US$199 a month at list price into any annual AI budget. For the full plan math, see our Ahrefs pricing guide.

The bottom line

Ahrefs' AI tools are built on the same strength as its SEO tools: a large dataset drawn from real searches. Their weakness is presentation. Prompt counts change from page to page, Grok data has stalled, and some indexes will move to paid add-ons. Use the included checks first. Then decide. Pay for the index only when you know what you will do with it.