The Headline Finding
UltraScout.ai grew from 0 to 190,500+ Copilot citations in 47 days (14 July – 30 August 2026). Across that period, 117 pages were cited across 400+ distinct queries, with a peak daily rate of 15,427 citations on 26 August. Crucially, this growth was achieved with no paid promotion, no link building campaign, and no advertising. All data is first-party, pulled directly from Bing Webmaster Tools AI Performance.
To our knowledge, this is the largest publicly documented AI citation growth dataset from a single domain. For context, Astiva AI published a 53-day study peaking at ~7K citations; AY Automate reported 43.9K over 3 months. This is 190K in 47 days.
Starting Point — What Existed Before 14 July
We need to be brutally honest about our baseline. We did not start from a strong position.
The site launched in late 2025 with roughly 100 pages of content. We had only 22 referring domains and 62 total backlinks — a very thin authority profile. We used a .ai TLD, which is a known Bing geo-targeting disadvantage. As of 13 July, we had zero Copilot citations.
The Approach
We did not run a step-by-step playbook to exploit Bing's algorithm. Instead, we categorised the work into fundamental areas.
3.1 Technical foundation
We resolved several infrastructure issues that were suppressing Bing's ability to correctly index and attribute the site. These included redirect hygiene, internal domain configuration, and explicit geo-targeting configuration for the .ai TLD.
3.2 Content principles
Every page followed a consistent structure: answer-first (opening with a direct answer in the first paragraph), question-style headings matching real user phrasing, original data points and proprietary frameworks, and complete JSON-LD Article schema with a canonical author entity.
3.3 Data-driven content expansion
Rather than publishing content at random, we performed a detailed gap analysis on our own UltraScout AI GEO Platform of the AI search landscape to identify exact citation gaps — queries where competitors were cited but we weren't. We then deepened existing high-performing pages rather than creating new ones from scratch, because a URL with existing Bing trust captures adjacent queries faster.
3.4 Indexing acceleration
Every content change was followed by an IndexNow submission to ensure Bing recrawled updated pages promptly. This is a standard, publicly documented protocol, and its consistent application materially affected recovery speed.
3.5 What we deliberately did NOT do
The Data — The Full 47-Day Timeline
The growth occurred in three distinct phases, each showing a unique pattern of explosive growth, brutal reset, and faster recovery.
Phase 1: Initial growth (14–26 July, 13 days)
Citations grew from 0 to a peak of 2,382 per day, with cited pages growing from 3 to 69. This happened with no content changes during that period, indicating Bing was finally recognising our entity structure.
Phase 2: Crash and recovery (27 July – 20 August, 25 days)
On 27 July, citations crashed from 2,063 to 52 overnight — a 99.7% drop. For about 11 days, we sat at near-zero. Then, on 7 August, we recovered to 6,649 citations in a single day, a 3x jump from Phase 1's peak. We sustained 8,000–13,000 citations per day, peaking at 13,202 on 17 August.
Phase 3: Current growth (21–27 August, ongoing)
On 21 August, we experienced a second crash down to 18 citations. But this time, recovery took only 3 days. By 25 August, we were at 14,088. By 26 August, we set a new all-time record of 15,427 citations, with 116 pages cited.
What Drove the Citations (The Content Analysis)
We analysed the pages driving this growth to understand what type of content earns citations.
5.1 The power law
The top 6 pages account for roughly 65% of all citations. The remaining 111 pages share the rest.
5.2 What earns citations vs. what doesn't
- Educational reference guides and pages with original data and frameworks earn thousands of citations — 3x-5x more than generic guides
- Product pages (pricing, features) earn near-zero citations
- Pages with answer-first structure consistently outperform those that bury the answer
- Content with inline CTAs also earns citations — publishers do not need to sacrifice conversion elements to win AI visibility
The Crash/Recovery Cycle — What We Learned
The crash/recovery cycle is not a penalty; it is a pipeline reset.
All pages dropped simultaneously (69 pages to 3 pages in one day), which points to a domain-level reset, not a page-level quality issue. No content changes preceded the crashes, and organic search rankings remained completely unchanged throughout.
What This Means for Other Brands
This study isn't just about us. It offers 8 practical, actionable takeaways for any brand trying to win AI search.
- 1Content structure matters more than domain authority for AI citations. We earned 190K citations with 22 referring domains.
- 2Answer-first structure is non-negotiable. Pages that bury the answer earn fewer citations.
- 3Original frameworks earn disproportionate citations. Proprietary concepts get cited because AI assistants can't find the same information elsewhere.
- 4The crash/recovery cycle is normal — don't panic. All pages drop and recover together. IndexNow submissions during the dead period position you for a stronger recovery.
- 5Publish frequency fights freshness decay. 2+ pieces per week maintains citation momentum. Long gaps correlate with drops.
- 6Product pages don't earn citations — guides do. Build authority through educational content first.
- 7Internal technical configuration matters. Fixing redirect hygiene, explicitly geo-targeting a
.aiTLD, and ensuring proper crawlability created the foundational infrastructure necessary for citations to compound. - 8.ai domains can earn citations despite the TLD disadvantage. Explicit geo-targeting and content authority overcome this.
We used the UltraScout GEO Platform to run the gap analysis behind this growth — identifying exactly which queries competitors were cited on and we weren't.
Methodology and Data Sources
For full transparency, all citation data was pulled from Bing Webmaster Tools → AI Performance (beta). Bing Webmaster Tools' Search Performance report was used for query and impression analysis. The data period covers 14 July – 30 August 2026 (47 days).
We acknowledge Bing's own disclaimer regarding the data: "data shown represents a sample of overall activity — results may be refined as additional data is processed." We have accounted for potential backfill effects in our analysis. No external audit was performed; this is first-party data, and readers should evaluate accordingly.
We note that Microsoft has acknowledged some citation data includes historical backfill. However, the consistent growth trajectory across 47 days — including two full crashes and recoveries — is inconsistent with a one-time backfill event: a backfill would appear as a single retroactive spike, not a pattern of daily accumulation, sudden resets to near-zero, and multi-day recoveries that each landed at a new high.
Frequently Asked Questions
How did UltraScout grow AI citations from 0 to 190,500 in 47 days?
Through content and technical foundation work only — no paid promotion, no link building, and no advertising. We resolved technical issues suppressing Bing's ability to index and attribute the site, published answer-first content with original data and frameworks, ran a data-driven gap analysis, deepened existing high-performing pages, and submitted every content change to IndexNow.
What caused the citation crashes during the growth period?
The exact cause is unconfirmed, but the pattern points to a domain-level pipeline reset rather than a quality penalty: citations dropped across all cited pages simultaneously in a single day, no content changes preceded either crash, and organic rankings remained unchanged throughout. Each crash recovered to a higher peak, with recovery time shrinking from 11 days to 3 days.
What type of content earns the most AI citations?
Educational reference guides and pages with original data or frameworks earned 3-5x more citations than generic guides. Answer-first structure outperformed content that buried the answer. Product and pricing pages earned near-zero citations. The top 6 pages accounted for roughly 65% of all citations.
Does IndexNow help AI citation recovery?
In this dataset, the fastest recovery phase directly correlated with content updates and IndexNow submissions fired during the dead period following a crash, producing a faster and stronger comeback than the previous cycle.
Can a .ai domain overcome Bing's geo-targeting disadvantage for AI citations?
Yes. This site uses a .ai TLD, a known Bing geo-targeting disadvantage, and started with only 22 referring domains and 62 total backlinks. Explicit geo-targeting configuration combined with content authority produced 190,500+ citations in 47 days.