Portfolio / AI search
AI Overview optimization
Priority service pages showed up for dozens of long, conversational searches but won almost no clicks. I added FAQs written in the exact words people were searching, so Google AI Overviews and other AI answers could quote them.
+31%AI Overview impressions on Atlassian Consulting Services
+29%AI Overview impressions on Jira Implementation Services
7service pages optimized
133queries grouped into 7 clusters on one page
The problem
The pages appeared for long, question-style searches but had no Q&A sections phrased the way people, and AI engines, were asking. They earned impressions, not clicks or citations.
What I did
- Pulled the real queries for each page from Search Console and grouped them into intent clusters. On the Atlassian Rovo page alone, 133 queries fell into 7 clusters.
- Added FAQ entries using the searchers’ own wording, without rewriting the rest of the page.
- Flagged cannibalization, intent mismatches, and title and click-through gaps for each page.
- Added AI Overview tracking to the weekly report, so every change could be measured week over week.
Result
One to two weeks after the edits, Google’s Generative AI features report showed (last 7 days compared with the 7 days before):
| Page | Before | After | Change |
|---|---|---|---|
| Atlassian Consulting Services | 147 | 193 | Up 31% |
| Jira Implementation Services | 63 | 81 | Up 29% |
| Atlassian Rovo | 50 | 44 | Down 12% |
| Atlassian AI Consulting | 21 | 10 | Down 52% |
Two pages showed a clear early lift. Two dropped, and I traced a likely cause: the AI Consulting and Rovo pages compete for the same positioning. Three more pages were edited too recently to judge, so I left them out rather than claim a result.
Testing visibility in AI assistants
I also tested 9 buyer prompts in ChatGPT, Perplexity, and Gemini and logged which sources each one cited. Perplexity named Empyra in 1 of 8 prompts, and Gemini cited Empyra’s pages for definitions in 2 of 7 without recommending it by name. That showed the content was trusted but not yet positioned to answer “who should I hire,” which shaped the next round of page changes.
What I learned
- Writing FAQs in the searcher’s exact words works better than rewriting whole pages.
- Pages that compete for the same topic can cancel each other out in AI answers.
- Report the drops alongside the wins; they show where to look next.
Honest noteThese are early signals measured one to two weeks after the edits, not a long-term trend. I included the two pages that dropped.
At a glance
- Up 31% and up 29% on two pages
- 2 pages dropped, cause traced
- 9 AI prompts tested
- Weekly AI Overview tracking
Skills
Tools
More case studies
Want more people to find your website?
I’m open to new projects and senior SEO roles in Bengaluru. Tell me about your site and I’ll reply with where I’d start.