Ahrefs Brand Radar July 2025 Update: New Indexes and AI Citations

Ahrefs added a 12 million-query Microsoft Copilot index, a Gemini index, and a 10x expansion of its ChatGPT and Perplexity data in its July 2025 release. This Ahrefs Brand Radar July 2025 update gave marketers a broader way to investigate AI search visibility, competitor appearances, and cited pages—but the useful payoff comes from knowing exactly what each dataset can show and what it cannot prove.

The release also brought AI-citation visibility into Site Explorer. Rather than treating every AI mention as equivalent, marketers could begin separating brand appearances from linked source pages and use both signals to identify content opportunities.

What changed in the July 2025 release

The July 2025 Ahrefs release included two new Brand Radar indexes and expanded coverage in existing AI-answer datasets. The central additions were a Microsoft Copilot index and a Gemini index, alongside more data for ChatGPT and Perplexity.

Ahrefs described the main AI-search changes as:

• A Microsoft Copilot index containing 12 million queries for July 2025.
• A new Gemini index tracking links from Google AI Overviews and AI Mode side by side.
• A 10x increase in the ChatGPT and Perplexity datasets.
• More long-tail questions sourced from Google’s People Also Ask data for ChatGPT and Perplexity.
• Citation data included for nearly every ChatGPT result in the index, according to Ahrefs’ announcement.
• AI citation reporting in Site Explorer, including visibility into pages appearing in AI responses.

This was primarily an expansion of available datasets and reporting surfaces. It should not be read as proof that every platform returns the same kinds of answers, uses the same source-selection process, or offers directly comparable visibility rates.

For practical analysis, keep the reporting period attached to every finding. A July 2025 index result is evidence about Ahrefs’ tracked query set and collected responses at that time; it is not a permanent ranking, an audience estimate, or a forecast of what an AI assistant will say later.

Ahrefs Brand Radar July 2025 update: the two new indexes

The two new indexes covered different AI-search environments. They are useful for competitor research, but neither should be treated as a universal score for all AI visibility.

Microsoft Copilot: 12 million queries

Ahrefs introduced a Microsoft Copilot index with 12 million July queries. That gave users a dedicated dataset for researching brands, cited domains, and recurring topics within Copilot responses.

A practitioner could use the Copilot index to investigate questions such as:

• Which brands are named for a product category?
• Which domains are linked when Copilot answers a research question?
• Which competitor names recur across a cluster of related prompts?
• Are certain product categories or use cases absent from the brand’s representation?

The index is valuable because it creates a consistent dataset for comparison. If one competitor appears repeatedly across a defined group of tracked Copilot queries while another does not, that is a useful research lead.

It does not, by itself, establish how many people viewed those answers, clicked a citation, or bought from a named company. It also does not prove that the same result will be returned for every formulation of a similar question.

Gemini: links from AI Overviews and AI Mode

Ahrefs’ announcement said the new Gemini index tracked links from Google AI Overviews and AI Mode side by side. That wording matters. The release described a reporting view of links from these Google AI experiences; it did not establish that they are identical products, use identical retrieval systems, or should be combined into one undifferentiated performance number.

For marketers, the practical benefit was the ability to inspect linked sources in two prominent Google AI experiences within the same index. A cited page may reveal an opportunity to study the source format, evidence, specificity, or topic coverage that appears in the tracked response.

A citation is still a narrow signal. It means a page or domain was linked in a collected AI response for a tracked query. It does not prove that the page ranks organically for the query, that it was the sole basis for the answer, or that it will be cited again in a future response.

The expanded ChatGPT and Perplexity datasets

Ahrefs said its ChatGPT and Perplexity datasets became 10 times larger in the July 2025 update. It also said the expansion included more long-tail questions drawn from Google’s People Also Ask data and new data providers.

That detail is useful because it describes the source of some added questions: People Also Ask can surface more specific, question-shaped searches than broad head terms. For example, a short query such as “project management software” and a longer question such as “what project management software works for a remote design agency” can produce very different AI answers.

The announcement does not establish that every prompt in every Brand Radar index was modeled on real searches. The defensible interpretation is narrower: Ahrefs expanded ChatGPT and Perplexity coverage with more long-tail People Also Ask questions, while maintaining separate datasets for the platforms it reported on.

Why a larger dataset changes comparisons

When an index expands, raw totals can be misleading. Suppose a company appears in 50 responses from a 1,000-prompt dataset, then appears in 100 responses after the dataset grows to 20,000 prompts. Its raw appearances doubled, but its appearance rate changed from 5% to 0.5%.

That is why reports should retain at least four fields:

1. The platform or index being measured.
2. The reporting period.
3. The number of tracked prompts or responses.
4. Both the raw count and the rate of appearances.

Without those fields, an increase can reflect broader coverage rather than stronger brand representation.

What Site Explorer added for AI citations

The July 2025 release added AI-citation visibility to Site Explorer. Ahrefs’ announcement positioned this around seeing how often top pages appear in AI responses across platforms, giving site owners a page-oriented way to investigate AI-search exposure.

That connection matters because Brand Radar-style reporting can identify an AI-answer pattern, while Site Explorer can help direct attention to the pages associated with that pattern. A domain-level view alone may tell a team that it is cited; a page-level view can help identify which resource, guide, report, or product page is appearing.

Consider a B2B software company with two pages:

• A research report about ransomware trends.
• A product comparison page for endpoint protection tools.

If the research report appears in AI responses about cybersecurity statistics while a competitor’s comparison page appears for “best endpoint protection platform,” the company has two different jobs to assess. One is maintaining evidence-led informational content; the other is improving decision-stage content for buyers comparing products.

Site Explorer reporting can support that investigation. It cannot demonstrate that a page caused an AI answer to mention a brand, or that the cited URL was the only source used in generating the response. AI answers may synthesize information from multiple linked and unlinked sources.

Mentions, citations, and share of answer are different metrics

AI visibility reporting becomes much more useful when teams avoid collapsing several signals into one number. The July 2025 update particularly highlighted citations, but a citation is not the same thing as a brand mention.

• Brand mention: The AI answer names a company, product, or brand.
• AI citation: The response includes a link to a page or domain.
• Page appearance: A specific URL is shown as a linked source in a tracked response.
• Share of answer: A comparative measure of how often one brand appears relative to named competitors in a defined set of answers or prompts.

For example, an answer could name “Brand A” as a recommended vendor but cite an independent review site and two documentation pages. Brand A has a mention, while the cited domains may belong to entirely different organizations.

Conversely, a company’s guide can be cited without the company being presented as a recommended solution. That may still be valuable content visibility, but it is not equivalent to winning a buyer comparison.

When calculating share of answer, define the answer set and brand-matching rules first. Decide whether product names, parent-company names, abbreviations, and common misspellings count. Then apply the same rules to every competitor. A share-of-answer result without a stated prompt set and matching method is difficult to interpret.

What the July 2025 data could support

Ahrefs’ new indexes and Site Explorer reporting gave marketers evidence to prioritize research. The strongest use is comparative: identify repeated patterns across a defined platform and query set, then inspect the actual pages and answers involved.

You could reasonably use the July 2025 data to:

• Find competitors named in tracked Copilot, ChatGPT, Perplexity, or Gemini-related responses.
• Identify domains and pages linked in tracked AI answers.
• Spot long-tail question themes where a competitor appears more consistently.
• Build a content-research queue from repeatedly cited sources.
• Compare your own site’s linked pages with competing pages addressing the same topic.

A repeated pattern is more informative than one isolated answer. If a competitor is linked across ten related queries about implementation, pricing, or product comparisons, that is a worthwhile signal to study. The next step is not to copy the competitor’s wording; it is to identify what evidence or task-specific information buyers may be missing from your own content.

What marketers should not conclude from the data

AI-answer monitoring is useful precisely because it reveals a changing surface. It becomes unreliable when teams turn a tracked appearance into a claim about traffic, causation, or universal platform preference.

The July 2025 release does not allow a marketer to conclude that:

• A citation generated visits, leads, or revenue.
• A named brand appears in every answer to the same question.
• A single cited page was the sole source behind an AI-generated response.
• More raw appearances necessarily mean a higher appearance rate after a dataset expansion.
• A cited page will remain cited in later responses.
• An AI response represents every buyer’s experience with a platform.

These constraints should shape reporting language. Say that a domain “appeared in tracked responses” rather than “ranked number one in AI.” Say that a competitor “was repeatedly named in this query set” rather than “dominates all AI search.” Precision protects both the analysis and the decisions that follow.

A practical workflow for turning citations into action

The most productive use of AI citation data is a narrow content investigation. Start with a cluster of commercially relevant questions, then review the cited pages and competing brands within that cluster.

A useful five-step workflow is:

1. Select a topic cluster, such as “best payroll software for nonprofits” or “how to implement endpoint protection.”
2. Record the platform, date, exact prompt, named brands, and linked domains for every tracked answer.
3. Group repeated citations by page type: documentation, original research, comparison page, review, product page, or editorial guide.
4. Compare your relevant page against the cited sources for directness, evidence, dates, product detail, and whether it actually answers the question.
5. Make one targeted improvement, then re-check the same prompt cluster rather than relying on a single new answer.

For example, if third-party implementation guides repeatedly appear while your own product documentation does not, the problem may be missing task-level guidance rather than insufficient homepage copy. If a competitor’s original survey is repeatedly cited, publishing generic commentary is unlikely to close the gap; the opportunity may require original data or clearer primary evidence.

Keep a change log. Record the page changed, what was added, the publication date, and the response pattern before and after the update. That does not prove causation, but it creates a much stronger basis for learning than reacting to dashboard totals alone.

Hosted indexes and selected-prompt monitoring

A large hosted index and a selected-prompt tracker address different needs. A hosted index such as Ahrefs Brand Radar is useful for discovery because its dataset can surface query themes, competitors, and cited sources a team would not think to monitor manually.

Selected-prompt monitoring is a general industry practice: a team chooses the questions that matter to its sales process, product category, support needs, or competitive positioning, then observes answers over time. Those prompts may come from sales calls, search-query research, support tickets, customer interviews, or competitor comparisons.

At AI Visibility Tracker, we use a local-first desktop workflow for the second job: monitoring a defined set of buyer questions and identifying brand mentions, citations, competitor gaps, and share of answer in those tracked answers. This is separate from Ahrefs’ July 2025 product release and should not be confused with a claim about Brand Radar’s platform coverage or collection method.

The trade-off is straightforward. Hosted indexes provide breadth and discovery; a carefully chosen prompt set provides relevance to the questions a specific business cares about. Neither is a complete replacement for the other. The stronger operating model uses broad data to discover themes, then validates important findings against a stable, explicitly documented set of buyer questions.

FAQ

What new features did Ahrefs release in July 2025 for Brand Radar?

Ahrefs added two Brand Radar indexes in July 2025: Microsoft Copilot and Gemini. The Copilot index contained 12 million July queries, while the Gemini index tracked links from Google AI Overviews and AI Mode side by side. Ahrefs also said it expanded ChatGPT and Perplexity data by 10x and added AI-citation visibility in Site Explorer.

What are the two new indexes in Brand Radar?

The two new July 2025 Brand Radar indexes were Microsoft Copilot and Gemini. The Copilot index gave Ahrefs users a dedicated set of 12 million July queries. The Gemini index tracked links from Google AI Overviews and AI Mode side by side. They represented separate AI-search environments and should be analyzed as separate datasets.

How does Site Explorer show AI citations and top pages appearing in AI responses?

The July 2025 Site Explorer update added AI-citation visibility intended to show how often top pages appeared in AI responses across platforms. This helps marketers move from a domain-level observation to page-level research: identify linked pages, inspect the relevant prompt and answer, and compare the page with competing sources that appear for the same topic.

Which AI platforms and prompts did Ahrefs Brand Radar cover in July 2025?

The July 2025 announcement discussed ChatGPT, Perplexity, Microsoft Copilot, and Gemini-related Google experiences, including AI Overviews and AI Mode. Ahrefs said ChatGPT and Perplexity data expanded with more long-tail questions from Google’s People Also Ask data. The announcement does not establish that every prompt in every index was sourced or modeled in the same way.

How can marketers use AI citations to improve brand visibility and content performance?

Use AI citations as a research signal. Review the exact prompt, answer, cited page type, and competing sources, then identify whether your content lacks direct answers, original evidence, task-specific documentation, or decision-stage information. Make a focused improvement and re-check a stable question set. A citation demonstrates appearance in a tracked response, not traffic or permanent placement.