Part 4: Page 1 on Google, Zero AI Citations
This post is Part 4 of four in a series about getting cited by AI search engines like ChatGPT and Perplexity.
- Part 1: Answer Capsules: The Content Trait LLMs Cite Most
- Part 2: How to Track Whether AI Is Citing You
- Part 3: How to Audit Any Page for AI Citability
We have a page on our sister site, BrassTranscripts, that ranks position 5.5 on Google for its target queries. It is indexed, well-structured, and getting organic traffic. But when we checked GA4 for AI referral traffic from ChatGPT, Perplexity, or Claude, the number was zero.
The 38% gap in action: Part 2 of this series cited research showing that 62% of brands on Google's page one appear in ChatGPT answers. This page is in the other 38% — ranking well in traditional search but invisible to AI.
This post is the case study we promised in Part 3. We ran the AI Citability report on a real page, made the recommended changes, and documented every step. This is what the process looks like in practice.
As with the rest of this series, this post uses the answer capsule technique throughout.
Quick Navigation
- The Page We Audited
- What the AI Citability Report Found
- What Was Already Working
- Fix 1: Adding Answer Capsules to Every H2
- Fix 2: Adding an Expert Citation
- Fix 3: Making the FAQ Section Visible
- The Before and After
- The Second Pass: Iterative Refinement
- What We Expect to Happen
- What This Means for Your Pages
- Frequently Asked Questions
The Page We Audited
We chose a law firm AI prompts guide on our sister site BrassTranscripts — a page ranking position 5.5 on Google with zero AI referral traffic.
The full page covers how law firms can use AI prompts to process depositions, client meetings, and case recordings.
The page was performing well by traditional SEO standards. It ranks in position 5.5 on Google — solidly on page one. Google last crawled it on March 6, 2026. It has clean heading hierarchy, front-loaded content, and specific data points (122 specialized prompts, a case study involving 102 audio files, processing speeds of 1-3 minutes per hour of audio).
By every traditional metric, this page is healthy. But it was getting zero traffic from AI systems. Not low traffic — zero.
What the AI Citability Report Found
The Brass-SEO AI Citability report identified three specific gaps preventing this page from being cited by LLMs, despite its strong Google ranking.
We ran the report using the Brass-SEO AI Citability button in Brass-SEO. It fetched the page, checked Google rankings and AI referral traffic, and evaluated all nine citability factors. (The audit has since added a tenth factor, scope clarity, after this case study was written.) Here is what it found:
| Factor | Assessment | Detail |
|---|---|---|
| Answer Capsules | 0 of 8 H2s | All headings led into lists or narrative without a standalone extractable sentence |
| Link-Free Capsules | 0 of 8 | No capsules existed to evaluate |
| Original Data | Found | "122 prompts", "102 audio files" case study, processing speeds |
| Brand Intent | Clear | Strong positioning statement in opening |
| FAQ Section | Missing in body | Schema markup existed but no visible Q&A text for LLMs to read |
| Content Front-Loading | Strong | First 30% clearly defined the value proposition |
| Expert Quotes | None | Entirely brand-led content, no external research or citations |
| Heading Hierarchy | Clean | Proper H1 → H2 structure |
| llms.txt | Found | Well-structured site-level file present |
Three factors needed fixing: answer capsules, expert citations, and the visible FAQ section. Three out of nine. The page was not fundamentally broken — it was missing specific structural elements that LLMs need to extract and cite content.
What Was Already Working
Four of the nine citability factors were already strong: original data, brand intent, content front-loading, and a site-level llms.txt file.
Original data was present throughout. The specific numbers — 122 prompts across four categories, a real case study involving 102 audio files, processing speeds of 1-3 minutes per hour — are exactly the kind of proprietary data that Part 1 of this series identified in 52.2% of cited blog posts. No other site has those exact numbers because no other site ran those exact tests.
Brand intent was clear. The page opened with a direct positioning statement about what BrassTranscripts is and what it does for law firms. An LLM encountering this page could immediately categorize it.
Content front-loading was strong. The key value proposition appeared in the first 30% of the text, which matters because 44% of LLM citations come from the first 30% of a page.
The llms.txt file was already in place at the site level, providing AI crawlers with a structured overview of the site's key content and services.
These factors did not need changing. The fixes were purely additive — we added to the page without removing anything that was already working.
Fix 1: Adding Answer Capsules to Every H2
We added a self-contained 1-2 sentence summary immediately after each of the 8 H2 headings. This was the highest-impact change.
Before the fix, every H2 on the page led directly into a list, a link, or a narrative paragraph. None had a standalone sentence that an LLM could extract as a citation. For example, the section "Legal Prompts for Case Preparation" opened with:
"Our prompt guide organizes 121 prompts across four categories. The Legal Professional category includes prompts for litigation support and case preparation."
That opening contains a link to the prompt guide and requires context from the heading to make sense. An LLM cannot quote it standalone. After the fix:
"Our legal AI toolkit provides 11 specialized prompts for deposition analysis, contradiction detection, and timeline construction to automate case preparation workflows."
The new capsule is self-contained — it answers "What are the legal prompts for case preparation?" without needing the heading, any links, or any additional context. This is exactly the pattern found in 72.4% of ChatGPT-cited blog posts.
We repeated this for all 8 H2 headings. Each capsule was 120-150 characters, link-free, and designed to stand alone as an extractable citation.
Fix 2: Adding an Expert Citation
We added a new section citing the Thomson Reuters 2025 Future of Professionals Report to provide the external authority signal the page was missing.
The Brass-SEO AI Citability report flagged that the page was "entirely brand-led content" — no external research, no expert quotes, no third-party data. Princeton's GEO research found that adding expert quotations improved AI visibility by up to 41%, making this the second-most impactful optimization technique tested.
The new section, "Why Law Firms Are Adopting AI Transcript Workflows," cited three specific findings from the Thomson Reuters report:
- 77% of lawyers use AI for document review
- 74% use it for legal research and document summarization
- AI adoption doubled from 14% to 26% between 2024 and 2025, with 78% expecting AI to become central to legal work within five years
These are verifiable statistics from a recognized authority in legal industry research. They give LLMs exactly what they need: credible third-party data that supports the page's thesis, with a direct link to the original source.
Fix 3: Making the FAQ Section Visible
We added 4 Q&A pairs as visible body text, matching the existing FAQ schema markup that LLMs could not see.
This fix addressed a subtle but important problem. The page already had FAQ schema markup — the structured data that tells Google "this page contains frequently asked questions." But that schema existed only in the HTML metadata. The actual questions and answers were not visible as formatted text in the body of the page.
Why does this matter? LLMs that crawl rendered HTML — which is how most retrieval-augmented generation systems work — could not see the FAQ content. They saw the body text, which had no Q&A section. The schema told Google about the FAQs, but the body text told LLMs there were none.
The fix added four questions to the body text:
- Can AI prompts be used for official legal proceedings?
- How do bulk transcription and AI prompts work together for law firms?
- Which AI tools work with these legal prompts?
- What does bulk transcription cost for a law firm processing case recordings?
As we covered in Part 3, Q&A is the highest-cited content format for LLMs because it mirrors the conversational query format that users type into AI systems.
The Before and After
Three fixes turned a 0-of-9 citability score into a page that meets all nine factors — in approximately 30 minutes of work.
| Factor | Before | After |
|---|---|---|
| Answer Capsules | 0 of 8 H2s | 8 of 8 H2s |
| Link-Free Capsules | N/A (none existed) | 8 of 8 link-free |
| Original Data | Present | Present (no change needed) |
| Brand Intent | Clear | Clear (no change needed) |
| FAQ Section | Schema only, not visible | 4 Q&A pairs in body text |
| Content Front-Loading | Strong | Strong (no change needed) |
| Expert Quotes | None | 1 (Thomson Reuters 2025 report) |
| Heading Hierarchy | Clean | Clean (no change needed) |
| llms.txt | Present | Present (no change needed) |
Total content added: approximately 350 words across capsules, the expert citation section, and the visible FAQ section. No content was removed. All changes were additive.
Time spent: approximately 30 minutes. Most of that was writing the 8 answer capsules. The expert citation required finding a credible report and summarizing its key findings. The FAQ section was the fastest — the questions already existed in the schema markup and just needed to be added as visible text.
The Second Pass: Iterative Refinement
The first AI Citability scan after our changes detected only 3 of 8 answer capsules and missed the expert citation entirely. A second round of targeted refinements resolved everything.
This is an important practical detail: AI citability optimization is iterative, not one-and-done. After making the initial changes described above, we ran the Brass-SEO AI Citability button on the updated page. The report showed that several capsules were not fully self-contained — they still required the heading for context or included qualifying language that made them harder for an LLM to extract standalone.
We refined the capsules to be tighter and more direct. We also reformatted the Thomson Reuters expert citation as a blockquote, which made the external authority signal more prominent and easier for the tool to detect.
One practical lesson: after deploying changes, we needed to clear the website's CDN cache before the AI Citability tool could detect the updates. The tool fetches the live page, so if the CDN is serving a stale cached version, the report will reflect the old content. If you make changes and the report does not reflect them, clear your cache and re-run.
After cache clearing and the second round of refinements, all nine citability factors showed as resolved.
What We Expect to Happen
It is too early to measure results. We will track three signals over the next 2-4 weeks and update this post with findings.
Signal 1: AI referral traffic. The page currently receives zero sessions from ChatGPT, Perplexity, or Claude. Any AI referral traffic after these changes would be a direct signal that the citability improvements are working. We will check this in GA4 using the method described in Part 2.
Signal 2: Manual query audit. We will run the page's target queries through ChatGPT and Perplexity to see whether the page appears in AI answers. The baseline is that it does not appear now despite ranking on page one.
Signal 3: Google position stability. The page ranks at position 5.5. These changes should not hurt that ranking — answer capsules and FAQ sections both improve traditional SEO as well. But we will monitor to confirm.
We are being transparent about what we do not know yet. The research from Part 1 shows that these structural patterns correlate with AI citations, but correlation is not causation, and a single page is not a statistically significant sample. This case study demonstrates the process, not a guaranteed outcome.
What This Means for Your Pages
If a page ranks well on Google but gets zero AI referral traffic, the problem is almost certainly structural — not a lack of authority or relevance.
This case study illustrates the exact pattern we described across the first three parts of this series:
- The page had authority — position 5.5 on Google, properly indexed, strong organic performance
- The page had good content — original data, clear brand intent, front-loaded value
- The page lacked extractability — no self-contained sentences for LLMs to quote, no external citations for credibility, no visible Q&A format
The fixes were not a rewrite. They were not a new content strategy. They were specific, targeted structural changes that took 30 minutes. That is the practical reality of AI citability optimization: most pages need refinement, not replacement.
You can run this same process on your own pages. Either manually audit the nine citability factors or use the Brass-SEO AI Citability button on your dashboard. Either way, start with your highest-ranking pages — they are already in the pool of content that LLMs consider. The structural fixes unlock the citations.
For the complete GEO implementation roadmap, including the research behind these nine citability factors, see the Generative Engine Optimization guide.
Frequently Asked Questions
How long does an AI citability audit take?
Using the Brass-SEO AI Citability button, the full audit runs in seconds — it fetches the page, checks Google rankings and AI referral traffic, and evaluates all nine factors. Doing the same audit manually takes 15-20 minutes per page.
How long until AI referral traffic appears after making changes?
LLMs incorporate new content faster than Google — days rather than weeks for retrieval-augmented systems like Perplexity. However, training-data-based models like ChatGPT's base model update on longer cycles. We expect to see initial signals within 2-4 weeks.
Will these changes hurt my Google rankings?
No. Answer capsules improve your chances of winning Google's featured snippets. FAQ sections generate FAQ rich snippets. Expert citations add E-E-A-T signals. All three changes benefit traditional SEO as well as AI citability.
Do I need to audit every page on my site?
No. Start with your top 10 pages by impressions — the ones already ranking well on Google. These have the highest chance of being picked up by LLMs because they are already in the candidate pool. Use your Google Search Console data to identify them.
What if my page is not on Google's first page?
Focus on traditional SEO first. 62% of ChatGPT citations come from Google's page-one results. Getting to page one is the single most effective thing you can do for AI visibility. Once you are there, the citability optimizations in this series help you convert that ranking into actual AI citations.
Is this case study statistically significant?
No. A single page is not a controlled experiment. This case study demonstrates the process and documents what we changed. The underlying research — 72.4% answer capsule correlation, 41% improvement from expert quotes — comes from larger studies cited in Part 1 and Part 3. We will update this post with results as they come in.
This is Part 4 of our LLM Visibility series. Part 1 covers the content traits LLMs cite most. Part 2 covers how to track your AI visibility. Part 3 covers how to audit any page for citability.
The page audited in this case study is AI Prompts for Law Firms: Transcript Workflows on BrassTranscripts. The expert citation added references the Thomson Reuters 2025 Future of Professionals Report. All before-and-after data in this post reflects actual changes made in March 2026.