Getting cited in AI answers: 0 to 206 brand mentions in four months
A second channel aimed at the sources AI models cite, run in parallel with the fix described in an earlier case study.
TL;DR
- Outreach placements that had zero AI-answer mentions in June grew to 206 mentions by October 3, the clearest signal coming from a one-week window before a data-collection change.
- Those placements alone account for 30% of all 689 AI-answer mentions for the brand, counting only the tracked placements out of the full outreach list.
- AI share of voice rose from 22.1% to 26.2% between June and September, while the category leader fell from 70.9% to 52.5%.
- The brand moved into a stable #2 of 5 against four competitors, the first time its position wasn't swinging month to month.
0 → 206
AI-answer mentions citing outreach placements, Jun to Oct
30%
Share of all 689 brand mentions that trace to a tracked placement
22.1% → 26.2%
AI share of voice, Jun vs. Sep average
#2 of 5
Rank in AI share of voice against four competitors
The shift: from rankings to citations
Traditional SEO optimizes for a ranking algorithm. A growing share of search now happens inside an AI answer instead, ChatGPT, Google AI Overviews, Google AI Mode, Gemini, Perplexity and Copilot all synthesize a response and cite sources, often without the user clicking through to any of them. Ranking first on Google doesn't guarantee a mention inside that synthesized answer. Getting cited does, and citation depends on which sites the model already treats as a source for a given topic.
This is the same AI writing SaaS behind an earlier case study on fixing a homepage vs. landing page cannibalization problem. That fix ran February to September 2026. A second channel started alongside it in June, building presence on the pages AI models cite.
The method
The process had three steps. First, monitor which domains AI models actually cite when answering prompts relevant to the product category, using Ahrefs Brand Radar across all six AI surfaces. Second, pursue placements on the domains that showed up as citation sources, the same outreach discipline as conventional link building, just retargeted at a different list. Third, track whether AI answers that go on to cite the new placements also mention the brand.
Placements landed across four independent newsletters, a consumer tech publication, an analytics-industry site and a national business outlet, all of which already appeared as cited sources in AI answers for category-relevant prompts before outreach began.
Results
Tracking only the placements named above: when the outreach channel started in June, AI answers citing those placements mentioned the brand zero times. The first mention tied to a placement appeared July 11. By October 3 there were 206, out of 689 total brand mentions across all AI surfaces, a 30% share from this channel alone.
The cleanest read on causation is a narrow window. Ahrefs expanded its prompt set on July 17, which raised mention counts for every brand in the set and makes totals before and after that date not directly comparable. In the six days before the expansion, July 11 to 16, total brand mentions rose from 122 to 190. Of those 68 new mentions, 37 came from answers citing a tracked placement, a window where the before and after numbers are actually apples to apples.
| Brand | Jun 2026 avg. | Sep 2026 avg. | Change |
|---|---|---|---|
| This brand | 22.1% | 26.2% | +4.1 pts |
| Competitor A (category leader) | 70.9% | 52.5% | −18.4 pts |
| Competitor B | 12.6% | 17.9% | +5.3 pts |
| Competitor C | 20.9% | 16.1% | −4.8 pts |
| Competitor D | 6.7% | 5.6% | −1.1 pts |
Source: Ahrefs Brand Radar, across ChatGPT, Google AI Overviews, Google AI Mode, Gemini, Perplexity and Copilot. Share of voice is the share of AI answers mentioning any brand in the competitive set that mention a given brand; one answer can name several brands, so the column doesn't sum to 100%.
Over the last 30 days tracked, the brand averaged 26.0% share of voice, second only to the category leader and ahead of Competitor B (18.0%), Competitor C (15.9%) and Competitor D (5.5%). That's a change in kind as much as degree: earlier in the year the brand's position swung between roughly 5% and 46% month to month as the category leader whipsawed between 50% and 80%. By the second half of the campaign, the ranking had settled into a consistent #2.
What this case illustrates
- 01
An AI answer cites sources
Ranking first on Google says nothing about whether an AI model treats a site as a source. The two lists overlap, but they aren't the same list, and the second one determines whether a brand gets mentioned inside a synthesized answer.
- 02
Outreach economics transfer, the target list doesn't
Identifying a target, pitching it, landing a placement: that's the same work as conventional link building. What changes is which sites qualify as targets. A domain Google barely ranks can still be a domain every AI model cites for a given topic.
- 03
Check the methodology before trusting a spike
The July 17 prompt-set expansion would have read as a huge jump in citations if taken at face value. Isolating the six days before it, July 11 to 16, gave a smaller but trustworthy number instead of a flattering but misleading one.
Disclaimer: All ideas and data are mine, Claude helped organize it and publish.