.png)
Somantra AI, an Enterprise-Grade AEO GEO and AI Search Visibility Platform, Publishes "Content Graveyard" Analysis: 57% of Domains Cited in ChatGPT and Google Are Never Cited Again - Yahoo Finance Singapore
TL;DR
We treat the 57.2% one-month citation claim as a reported result from a defined Australian insurance dataset, not a universal AI visibility benchmark. We explain what the announcement established, what it cannot establish, and how marketing teams can measure citations, mentions, and content gaps across priority prompts.
Somantra AI, an Enterprise-Grade AEO GEO and AI Search Visibility Platform, Publishes "Content Graveyard" Analysis: 57% of Domains Cited in ChatGPT and Google Are Never Cited Again - Yahoo Finance Singapore
AI answers are no longer a peripheral research channel for marketing teams. In a March 2025 browsing study, 58% of U.S. adults encountered a Google search page with an AI-generated summary.
On 3 August 2026, Somantra published an Australian insurance citation study reporting that 57.2% of tracked domains appeared in only one observed month. For AI visibility teams, this is a publisher-reported warning, not a universal benchmark: measure priority prompts repeatedly across platforms, record cited URLs and context, and change content only when patterns persist.
We separate the confirmed announcement from its unreplicated conclusions, explain its reported method, and turn the result into a practical AEO measurement workflow.
What Happened on 3 August 2026
The reported event is straightforward. A Yahoo publication announced a new Australian insurance citation analysis, while also labelling the item as a paid press release. That label matters because it distinguishes confirmation that the report was published from independent verification of its findings.
The report says it analyzed citation records from ChatGPT search and Google across seven observed months. Its headline result was that 57.2% of tracked domains appeared in one observed month and did not reappear in the others. The useful editorial response is neither to dismiss that claim nor treat it as settled market truth. It is to understand the reported evidence, then test whether the same pattern exists for the prompts, countries, and buyer journeys that matter to your business.
| Item | Reported Detail | Editorial Treatment |
|---|---|---|
| Publication date | 3 August 2026 | Confirmed event date |
| Study scope | Australian insurance category | Do not generalize automatically |
| Headline result | 57.2% cited in one observed month only | Publisher-reported finding |
| Persistent domains | 279 domains, or 2.7%, appeared in all seven months | Dataset-specific observation |
A citation can be meaningful without being durable. The right next step is to preserve source evidence, define the observation method before running it, and compare repeated checks instead of drawing a conclusion from a celebratory screenshot. Our citation tracking method starts with reproducible evidence rather than a single visibility score.
What the Content Graveyard Result Actually Measures
The phrase “content graveyard” is memorable, but the underlying claim is narrower than the phrase suggests. It does not show that a page is permanently invisible, that a brand has failed at content, or that every AI answer surface behaves alike. It describes domain appearances within one publisher’s defined sample.
The Reported Dataset Was Large but Specific
According to the release methodology, the analysis contained 2,437,107 citation records across 28,725 unique domains. The reported observation months were November and December 2025, then January, February, March, May, and July 2026.
Each cited record reportedly contained a month, platform, domain, URL, position, title, and description. That is enough information to assess recurring domain appearances, but it is not the same as a continuous daily study of citation volatility. The missing months and changing query sample sizes mean raw monthly totals should not be read as a clean measure of brand momentum.
The 57.2% Finding Describes Domain Persistence
The study reports that 57.2% of domains appeared in a single observed month, while 279 domains, or 2.7%, appeared in every observed month. Those figures are useful because they shift the conversation from “Were we cited?” to “Did we remain visible when the same category was observed again?”
The most practical implication is to make persistence a metric. A one-time citation may indicate that a page was eligible for an answer. Repeated inclusion is stronger evidence that the page, topic, and source profile continue to fit a recurring answer pattern.
The Platform Split Should Not Become a Single Score
The report says 5,522 domains appeared on both platforms, while 18,279 appeared only in Google results and 4,924 only in ChatGPT results. It also reports 346,172 ChatGPT records and 2,090,935 Google records.
Those totals are not a reason to declare one platform easier or harder. They are a reason to avoid blended visibility scores that hide which prompt, platform, and page produced an appearance. A proper comparison holds the prompt and locale steady, retains the original answer evidence, and shows the platform-specific result before calculating any rollup. Our cross-engine method keeps those observations separate before comparison.
Format Associations Are Leads, Not Laws
The release reports that discount or savings language appeared twice as often among domains that persisted across all observed months, while comparison content appeared 1.9 times as often and how-to content 1.5 times as often. It also reports weaker persistence for guide-format pages.
These are associations inside a particular category and sample. They can inform a content audit, but they cannot prove that changing a heading or adding a format label will create durable citations. Intent, source quality, crawlability, topical usefulness, and the prompt itself still matter. Treat these observations as candidates for testing, then evaluate changes against a documented baseline rather than relying on a generalized content formula.
What This Means for AI Visibility
AI visibility is not a new substitute for SEO. It is the discipline of observing whether a defined set of buyer prompts produces a brand mention, a cited source, or a useful page appearance over time. The work becomes valuable when it connects those observations to editorial decisions.
Google’s own site-owner guidance is refreshingly direct: core SEO best practices remain relevant for AI Overviews and AI Mode, and there are no additional requirements or special optimizations required for inclusion. That means answer engine optimisation should not become a hunt for secret formatting tricks.
The practical change is measurement. Traditional rankings tell us where a result appeared in a list. Answer-led search can also show which source was used to support a response, what language surrounded it, and whether the same source returns on another run. A disciplined AI visibility system treats those as separate observations rather than forcing them into one number.
A cited page is not automatically a recommendation. A source may be linked as background, as a supporting fact, or as one perspective among several. Teams should therefore track answer context alongside presence, including whether the answer names the brand, whether the cited page directly supports the claim, and whether the answer creates a useful path to the next buyer question.
This is also where content, SEO, and product marketing need the same evidence. Content teams may see an opportunity for a clearer explanation. Search teams may see a discoverability issue. Product marketers may see vague positioning or missing proof. Our brand mention monitoring separates simple mention frequency from the language and source evidence that give a mention meaning.
| Measurement Question | Useful Evidence | Weak Shortcut to Avoid |
|---|---|---|
| Did we appear? | Prompt-level mention or cited URL | One blended score |
| Did we return? | Repeated observations using the same method | One screenshot |
| Did the right page appear? | Cited URL matched to intent | Counting only domains |
| Did the answer help buyers? | Context, wording, and next-step relevance | Treating every citation as endorsement |
How We Would Measure the Claim in a Real Content Program
A useful response to the announcement is not to refresh every old page. It is to build a compact measurement system that can reveal whether a content gap is real, repeated, and commercially relevant. Start with a prompt set that reflects how buyers actually ask for help, then keep the evidence stable enough to compare over time.
Build a Decision-Linked Prompt Set
Group prompts by the decision they support: defining a problem, comparing approaches, evaluating requirements, or selecting a provider. Capture the buyer role, market, language, and intended page for each prompt. Document why the prompt belongs in the set and what a useful answer should help the buyer understand.
Avoid prompts chosen only because they sound impressive in a report. A prompt should connect to a customer question, a content decision, or a positioning risk. That connection makes later findings actionable. If the answer varies, the team can still determine whether the variation affects a meaningful topic rather than treating every source change as equally important.
Start with a small, defensible set and add prompts only when the team can explain their business relevance. A structured buyer prompt dataset prevents a monitoring program from becoming a random collection of interesting questions.
Capture Evidence at the URL and Context Level
For every observation, record the date, platform, locale, exact prompt, full answer, source URLs, brand mention, cited-page context, and any meaningful answer change. Save the raw evidence before turning it into a dashboard. A dashboard is a view of the evidence, not a replacement for it.
ChatGPT search can display inline citations or a sources panel, as OpenAI explains. That makes it possible to distinguish a cited page from a brand mentioned without a source link. This distinction is essential when reviewing visibility because a mention can be positive, neutral, partial, or unsupported by a source the reader can inspect.
Context also creates editorial precision. If an answer repeatedly uses a brand as an example but cites another source for the key claim, the work may be to improve proof and source eligibility, not merely to increase mention counts. Our citation-context framework is designed around that difference.
Compare Like with Like Across Platforms
Keep the prompt wording, market, device assumptions, and run cadence consistent where possible. Then compare platforms separately before making an aggregate judgment. A page that appears in one surface but not another is a finding to investigate, not a data-cleaning problem.
When differences appear, examine what the answer asked for and how each response framed the topic. One answer may prioritize explanation, another may prioritize comparison, and another may cite a source for a narrow factual statement. The useful question is not which platform “won.” It is whether the business has a credible, accessible answer for the specific intent that produced the gap.
Compare source domains and exact URLs, then look for recurring themes across multiple runs. This approach avoids overreacting to a single output while preserving useful evidence about eligibility. Use cross-engine answer tracking to identify recurring source patterns before prioritizing work.
Turn Repeatable Gaps into Content Work
Prioritize gaps that recur for valuable prompts. Check whether the best existing page is crawlable, internally linked, easy to read in text form, and directly useful for the question. Then decide whether the right intervention is a refresh, a new answer page, stronger evidence, or a clearer internal connection.
A strong decision record should identify the problem before proposing the format. For example, a missing cited page may reflect outdated information, unclear definitions, weak first-party explanation, or a page that does not address the buyer’s actual question. Each diagnosis calls for a different response, and not all responses require creating more content.
After publishing or revising, rerun the same prompt set using the original method. Compare the evidence to the previous observation, note meaningful changes, and allow enough time to distinguish a persistent signal from normal variation. A content optimization stack helps keep measurement connected to accountable editorial decisions.
Put PageLens.ai to Work
If your team is moving from occasional screenshots to a governed AI visibility program, PageLens.ai is the practical place to start. We help marketing, growth, SEO, and content leaders turn a defined prompt set into evidence they can review: where a brand appears, which URLs are cited, how answer language changes, and which gaps deserve editorial work. The value is not a vanity score. It is a shared operating record that makes conversations between content, search, product marketing, and leadership more concrete. Begin with a small set of decision-linked prompts, agree on the evidence each team needs, and review movement on a repeatable schedule. That approach makes it easier to distinguish a one-off mention from a pattern that merits investment. When a persistent gap appears, we can help your team connect the finding to pages, content briefs, and accountable next actions. Book a demo
FAQs on AI Visibility
Is AI Visibility the Same as SEO?
SEO supports crawlability and ranking. AI visibility shows whether priority buyer prompts repeatedly produce brand mentions, cited pages, and useful context across answer-led search experiences over time.
Should We Treat the 57% Result as an Industry Benchmark?
No. It is a publisher-reported result from a defined insurance dataset. Use it to motivate consistent measurement of your own prompts, markets, cited pages, and platforms.
How Often Should We Measure AI Visibility?
Choose a repeatable weekly or monthly cadence that matches your buying cycle. Preserve prompt, locale, platform, and evidence fields so each comparison remains meaningful over time.
