The way users discover information is changing. For years, organic visibility has been measured primarily through rankings, impressions, clicks and positions in traditional search engines. However, with the arrival of AI-generated responses in tools like Microsoft Copilot and Bing, a new question emerges for website owners, publishers and marketing teams:
Is my content being used as a source in AI-generated responses?
To begin answering this question, Bing Webmaster Tools has introduced a new section called AI Performance — a set of metrics designed to show how a website’s content appears within AI experiences such as Copilot, Bing-generated summaries and certain partner integrations.
This change is significant because it extends traditional SEO analysis into a new dimension: visibility in generative engines, also known as GEO, or Generative Engine Optimization. It is closely related to modern AI SEO strategy.
What is AI Performance in Bing Webmaster Tools
AI Performance is a panel inside Bing Webmaster Tools that helps you understand how your website is being cited or referenced in AI-generated responses.
Until now, Bing Webmaster Tools mainly helped analyse aspects such as indexing, crawling, search performance, clicks, impressions, sitemaps and technical errors. With AI Performance, Bing adds a new layer: the participation of your content inside generative responses.
This means it is no longer only about whether a URL appears as a blue link in Bing, but also whether that URL is being used as a source when an AI generates a response.
Why this matters
In traditional search results, the user typically sees a list of links and decides which one to click. In a generative experience, the AI system can construct a response directly and cite several sources.
This changes how visibility is measured. A page may not receive an immediate click, but it can be used as a source to answer questions. That means the content participates in the response process, even though user behaviour is different from a classic SERP.
That is why AI Performance metrics help answer questions like:
- Which URLs from my site are being cited by the AI?
- What topics cause Bing/Copilot to use my content as a source?
- Which pages appear most frequently in generative responses?
- Is AI visibility growing or declining over time?
- Which content should I improve to be clearer, more complete and more useful?
What the AI Performance panel measures
The AI Performance dashboard includes several metrics designed to understand a website’s presence in AI responses.
1. Total Citations
This metric shows the total number of citations a site has received in AI-generated responses during a selected period. In other words, it shows how many times the site’s content appeared as a source.
It is important to understand that this metric does not necessarily indicate the exact position of the citation within a response or its visual prominence. It simply measures how many times the content was referenced.
2. Average Cited Pages
This metric shows the daily average of unique pages from the site that were cited in AI-generated responses. It helps you understand how many different pages are participating as sources.
For example, if a site has many citations concentrated on a single URL, the analysis will differ from a site where many different pages are being cited. This metric should not be interpreted as an authority ranking or as a measurement of individual page importance. It is an aggregate signal of the site’s participation in AI experiences.
3. Grounding Queries
Grounding Queries are one of the most interesting parts of the report. They represent phrases or queries that the AI system used to retrieve information that ended up being cited in a response.
They are not necessarily the exact prompts written by users. They are more like signals from the information retrieval system: topics, phrases or internal queries that helped the AI find relevant content.
For example, a Grounding Query may indicate that a website’s content is being used to answer topics such as definitions, comparisons, guides, local information, frequently asked questions, educational explanations, recommendations or specialised content.
For GEO analysis, this metric is key because it helps understand which topics the AI considers your content useful for.
4. Page-level citation activity
This section shows which specific URLs from a site have been cited and how many citations they have received. It is a very useful metric because it allows you to go from general analysis down to the page level.
With it you can identify which URLs are cited most, which content has the greatest presence in AI, which informational pages are being used as sources, which commercial pages have little or high presence and which topics could be better optimised.
That said: a page with many citations does not automatically mean it is «more important» or that it holds a higher position within a response. The metric indicates citation frequency, not ranking.
5. Visibility trends over time
The dashboard also shows how citation activity evolves over time. This allows you to detect trends such as increasing citations, declining visibility, one-off spikes, the impact of content updates, the effect of technical changes and behaviour before and after a migration.
This temporal view can be especially useful for analysing whether a content strategy is gaining traction within generative responses.
AI Performance and the birth of a new GEO layer
AI Performance represents an important step towards more specific Generative Engine Optimization tools. Traditional SEO has focused on improving visibility in search engines. GEO extends that logic to generative engines, where the goal is not only to appear in a list of results, but to be a useful, clear and reliable source for AI-generated responses.
This does not mean that SEO stops mattering. On the contrary, many fundamentals remain essential: crawlable content, clear architecture, topical authority, good indexing, useful content, up-to-date information and solid semantic structure.
The difference is that we must now also ask ourselves whether content is prepared to be understood, retrieved and cited by AI systems.
How to use AI Performance to improve content
AI Performance data should not be analysed only as a passive report. It can be used to make content decisions.
1. Identify pages already working in AI
If a URL appears cited many times, it is a signal that Bing/Copilot already considers it useful for certain topics. In that case, it is worth reviewing whether the content is up to date, whether it clearly answers user intent, whether it has internal links to relevant pages, whether it could be expanded with more examples and whether it is connected to business objectives.
A highly cited informational page can become an excellent entry point towards more commercial content.
2. Detect topics with potential
Grounding Queries allow you to identify topics where the site already has some visibility. If many queries are related to the same topic, it may be a good idea to create or reinforce a content cluster around that theme.
For example: if many queries are about definitions, improve glossaries; if comparative queries appear, create tables and guides; if local queries appear, strengthen city or area pages; if common questions appear, create more comprehensive FAQs.
3. Improve structure and clarity
AI systems need to understand quickly what a page is about. This is why it is advisable to work on elements like clear titles, well-organised headings, lists, tables, direct answers, FAQ sections, simple definitions, up-to-date data and a logical structure.
A clear page not only helps the user, it also makes it easier for retrieval systems to identify useful fragments. This aligns closely with technical SEO best practices.
4. Add evidence and context
When a page contains important claims, it is worth supporting them with data, examples, sources or explanations. This helps build trust and can improve the content’s usefulness as a reference.
Some useful elements are: real examples, figures, dates, step-by-step explanations, use cases, comparisons, references to official sources and additional context to avoid ambiguity.
5. Keep content up to date
Content freshness is especially important when AI can use a page as a source. If a URL contains outdated information, it may lose usefulness or generate less accurate responses.
It is worth periodically reviewing dates, prices, requirements, schedules, processes, legal data, product information, local information and service availability.
The importance of IndexNow
IndexNow is a protocol that allows you to notify search engines when a URL is created, updated or deleted. This is relevant because, if search and AI systems depend on up-to-date content, accelerating the discovery of changes can help the most recent version of a page be considered sooner.
It does not replace a good SEO strategy, but it can be useful for improving the speed at which engines detect important changes. On sites where content changes frequently, IndexNow can be an important piece of an update and maintenance strategy.
AI Performance for local businesses
For local businesses, AI visibility can also depend on basic data such as address, opening hours, phone number, services, location, business category, availability and contact information.
That is why, in addition to working on the website, it is important to keep information up to date in tools such as Bing Places for Business. When an AI answers local queries, it needs consistent and reliable data. If the business information is incomplete or outdated, it can reduce the likelihood of appearing correctly in location-related responses. Our Local SEO service covers exactly this kind of visibility.
What limitations does AI Performance have
Although AI Performance is a very useful tool, it is important to understand its limits.
It does not show all real prompts. Grounding Queries are not necessarily the exact user prompts. They are retrieval signals used by the system.
It does not measure clicks from AI responses. The dashboard helps understand citations and referenced pages, but does not always allow you to know whether those citations resulted in traffic.
It does not indicate exact ranking within a response. A frequently cited URL does not necessarily appear in a prominent position within each response.
Data may be a sample. As with many reporting tools, some data may represent a sample of total activity.
GEO methodology is still evolving. There is not yet a single definitive methodology for optimising content for generative responses. AI Performance is an important signal, but it must be combined with SEO analysis, content, structured data, logs, analytics and manual review.
How to interpret the data correctly
To get value from the dashboard, avoid an overly simplistic reading. It is not enough to look at which page has the most citations. You need to cross-reference several questions:
- Is the cited page informational or commercial?
- Does it have organic traffic from traditional search engines?
- Is it up to date?
- Does it have internal links to relevant pages?
- Does it answer search intent well?
- Does it have a clear structure?
- Is it aligned with business objectives?
- Are the Grounding Queries related to strategic topics?
This approach allows you to move from «this page appears in AI» to «this page can help build authority, traffic or conversion».
Practical analysis example
Imagine an educational website that detects its most cited AI pages are blog articles about grammar, culture and vocabulary. This may indicate the site has informational authority, but not necessarily that its commercial pages are gaining AI visibility.
The opportunity lies in connecting both worlds: adding internal links from informational articles to course pages, creating related modules, adding contextual calls to action, improving commercial pages with FAQs, answering questions the AI can use as a source and reinforcing comparison and decision-oriented guides.
In this way, informational visibility can support commercial objectives.
Which metrics to review together
For a more complete analysis, AI Performance should be combined with other sources.
From Bing Webmaster Tools
- Total Citations.
- Average Cited Pages.
- Grounding Queries.
- Cited URLs.
- Time trend.
From Google Search Console
- Clicks.
- Impressions.
- CTR.
- Average position.
- Organic queries.
- Pages with the most traffic.
From web analytics
- Sessions.
- Conversions.
- Forms.
- Scroll.
- Time on page.
- Key events.
From content analysis
- Content quality.
- Update frequency.
- Structure.
- Internal linking.
- Topical coverage.
- Answer clarity.
Combining these sources lets you understand whether a page only has visibility, whether it also generates traffic, and whether it also contributes to the business.
Best practices to improve presence in AI responses
Although GEO is still evolving, some practices make clear sense:
- Create clear, comprehensive content.
- Use descriptive headings.
- Answer questions directly.
- Add useful FAQs.
- Include tables when they help compare information.
- Keep content up to date.
- Cite sources or provide evidence where necessary.
- Avoid ambiguity in names, products and entities.
- Improve internal linking.
- Align informational content with conversion pages.
- Use structured data where applicable.
- Review which URLs are already being cited.
- Expand content that is already showing traction.
- Correct outdated content.
- Maintain basic technical signals: indexing, crawling, canonicals, robots.txt and sitemaps.
AI Performance does not replace traditional SEO
It is important to stress this: AI Performance does not replace Google Search Console, Bing Search Performance or web analytics. It is an additional layer.
Traditional SEO remains necessary to understand which pages rank, which queries generate traffic, which content has impressions, which URLs are losing or gaining visibility and which pages convert.
AI Performance adds another question: what content is being used as a source in AI-generated responses? The answer to that question can change how content, updates and opportunities are prioritised.
Conclusion
AI Performance in Bing Webmaster Tools is one of the first important steps towards greater transparency about how websites participate in AI-generated responses.
It lets you see which content is being cited, which pages appear as sources and which topics are connecting with generative experiences. Although it does not yet offer a perfect view of all prompts, clicks or positions within AI responses, it does provide very valuable signals for understanding a website’s GEO visibility.
For publishers, businesses and marketing teams, this opens a new form of analysis: not only measuring whether a page ranks, but whether it is also considered a useful source by artificial intelligence systems.
The opportunity lies in using this data to improve content, reinforce authority, keep information up to date and better connect informational content with the site’s real objectives.
In an environment where search is no longer limited to blue links, understanding AI citations is becoming an increasingly important part of digital strategy.
