What Gets Cited by Bing AI: Data from 16,000 Citations on SmallBizAI.au
In mid-March 2026, Bing AI was citing pages on SmallBizAI.au around 16 times a day. By May 18, that number hit 709 — across 61 different pages — in a single day.
That’s a 43x increase in 10 weeks. We now have richer data: 16,000 total citations, grounding query logs showing what people actually asked Bing Copilot, and a clearer picture of exactly which content earns AI trust.
Here’s what the full dataset shows.
The growth curve
Bing Webmaster Tools tracks AI citations separately from regular search. Here’s how the numbers moved from launch through to late May:
| Period | Daily citations (approx.) | Pages cited |
|---|---|---|
| Mid-March (launch) | 10–50 | 3–11 |
| Early April | 30–80 | 8–18 |
| Mid-April (from 13th) | 98–231 | 26–51 |
| Late April | 300–485 | 42–72 |
| May (sustained) | 400–709 | 41–71 |
Monthly totals: March (14 tracked days): 307. April: 4,582. May 1–24: 11,131. Grand total: over 16,000 citations in 10 weeks.
The first inflection point was April 13–14, when citations jumped from 77 to 214 overnight. The second is May itself: the site has settled into a sustained 400–700/day range, not a spike. That’s compounding — more pages, more queries, more surface area for Copilot to reach.
What’s actually getting cited
Across all pages cited over the full period, the top 10 dominate:
Seven of the top ten are head-to-head comparisons or detailed company profiles. The pattern from April holds: Bing AI reaches for content structured around a decision, not content that describes a category.
The Zeller pattern
Zeller’s profile page has attracted over 20 distinct grounding query variants — all citing the same page. A sample:
- “Zeller fintech company Square AI evaluation”
- “Zeller fintech company business banking Australia features pricing reviews”
- “Zeller fintech company cash flow management evaluation”
- “Zeller fintech company loyalty software capabilities”
- “Zeller fintech company contactless payments Australia”
Each is a slightly different question. Each gets the same answer. That’s what it looks like when a page earns AI authority on a topic: Copilot stops looking elsewhere and routes every related query back to the same source. One well-written, detailed profile can anchor an entire topic cluster.
The queries confirm it
The grounding queries file shows what people actually asked Copilot before it cited a page. The language is revealing:
- “how to use ai to handle complaints politely” — 660 citations from a single behavioural query
- “Tanda vs Deputy Australia hospitality retail shift rostering time tracking Modern Award compliance” — 459
- “Workday Bamboo Australia ME SMB evaluation” — 134
- “Firmable B2B contact database Australia ANZ data quality” — 130
- “best agentic AI solutions for small businesses Australia 2026” — 105
- “calendar scheduling automation CRM vs standalone apps Australian SMBs” — 14
These are evaluation queries — someone asking Copilot to help them choose or solve a specific problem. The content that gets cited is content that answers that decision directly, with enough specificity to be useful.
The complaints how-to is the standout single-query performer. 660 citations from one behavioural question: “how to use ai to handle complaints politely.” That’s not a brand query and it’s not a comparison — it’s a task query. People asking Copilot to help them do something, and Copilot reaching for a page that shows them how.
What the data actually shows
- Comparison content is what AI search reaches for. Head-to-head posts that name specific tools, give real differences, and take a position are exactly what Copilot serves when someone asks “which is better for Australian small business?”
- Depth on a single company anchors topic clusters. The Flare HR and Zeller profiles generate more citations than most comparison posts. When someone asks about these companies, Copilot has one page it trusts — and routes every variant of the question there.
- Behavioural how-to content converts differently. The complaints post (658 citations) and invoicing post (76) aren’t comparison content — they answer task queries. People asking Copilot to help them do something specific. This is a distinct citation path worth building for.
- Breadth amplifies volume. The May sustained run has 50–70 different pages cited per day. That’s not one great page — it’s many pages, each answering a narrow question well. Total surface area matters as much as individual page quality.
- Pricing content converts reliably. “How much does ChatGPT cost in Australia” (464 citations), “does ChatGPT have GST” (21 query citations). Practical cost questions are consistent citation drivers regardless of content type.
- Compounding is real. The site went from 307 citations in its first two weeks to 11,131 in the first 24 days of May. The content wasn’t rewritten. More pages, covering more queries, created more surface area. Citations compound.
📬 Get the SmallBizAI.au Weekly
Practical AI news for Australian small businesses — every week, free.
📬 The SmallBizAI Brief
One practical AI tip for Australian small business. Every Tuesday. Free.
Join business owners getting smarter about AI every week.
Sources
- Bing Webmaster Tools, AI Performance reports (AIPageStatsReport, AISearchQueriesReport, and AIPerformanceOverviewStats exports, March 18 – May 24, 2026)
📊 Compare AI tools side by side | 💼 Free resources & AI prompt packs
