How Odyssiant is improving its own AI visibility over time
An ongoing case study in how Odyssiant uses its own platform to measure, improve and retest its visibility across AI-generated answers.
Most marketing platforms make big claims about visibility. We wanted to test ours on ourselves. So we ran Odyssiant on Odyssiant — not as a neat demo account, not as a polished sales example, but as a live audit of our own AI visibility: where we appeared, where we were missing, how we were described, which competitors appeared around us, and what actions we needed to take.
Updated 31 May 2026
We have updated this case study with a later four-engine audit. The original one-month comparison remains in place, but the latest data shows the same direction of travel: Odyssiant's visibility continued to improve, even under a broader and harder test.
31 May update: visibility continued to improve under a broader test
Since this case study was first published, we have run a further Odyssiant audit on Odyssiant.
The latest run was not an easier test. It expanded the audit from one engine to four, increased the answer set from 105 to 520, and included additional prompts that reflect the way we now describe and test the AI visibility category. Even with that broader test, the overall score continued to move in the right direction.
| Run date | Engines | Answers | Overall score |
|---|---|---|---|
| 10 Apr | 1 | 105 | 17 |
| 28 Apr | 1 | 105 | 19 |
| 11 May | 1 | 105 | 23 |
| 31 May | 4 | 520 | 26 |
The cleanest like-for-like comparison is the OpenAI prompt set that existed across all four runs. On that comparable set, Odyssiant's average score improved from 17.4 on 10 April to 27.2 on 31 May.
| Run date | Comparable OpenAI average score |
|---|---|
| 10 Apr | 17.4 |
| 28 Apr | 19.2 |
| 11 May | 23.5 |
| 31 May | 27.2 |
Across that comparable OpenAI set, 38 prompts improved, 8 declined and 59 stayed the same. That gives us a more useful view than the headline score alone. The movement was not created by one isolated result. More prompts improved than declined, and the score continued to rise when we retested.
The latest four-engine run also added a set of new prompts. These performed above the existing prompt set, suggesting Odyssiant is becoming more visible against the commercially specific questions and language now being used to describe the category.
| Prompt type | Answers | Avg score |
|---|---|---|
| Existing prompts | 420 | 24.3 |
| New prompts | 100 | 34.1 |
The broader four-engine run also showed where the next work is needed. OpenAI, Gemini and Perplexity clustered around the high 20s, while Claude remained weaker.
| Engine | Avg score | Citation coverage |
|---|---|---|
| OpenAI GPT-4o | 28.7 | 99% |
| Google Gemini 2 Flash | 28.4 | 35% |
| Perplexity Sonar | 27.7 | 99% |
| Claude 4 Sonnet | 19.9 | 51% |
Evidence and citation coverage also became more nuanced in the four-engine test. OpenAI citation coverage stayed very high at 99%, and Perplexity also reached 99%. Gemini and Claude returned fewer cited answers, at 35% and 51% respectively. So the apparent evidence drop is not simply an Odyssiant evidence problem. It reflects how differently engines behave.
The stage pattern remained clear. Odyssiant is strongest when buyers are evaluating tools, comparing options and making decisions. Awareness remains the weakest stage.
| Stage | Latest avg score |
|---|---|
| Awareness | 0.0 |
| Consideration | 28.8 |
| Evaluation | 52.0 |
| Decision | 32.0 |
The lesson has not changed. AI visibility does not improve evenly. Product and evaluation-stage understanding can move with clearer content, better internal linking and stronger explanation. Broader awareness needs a wider evidence base: third-party citations, backlinks, category coverage, comparisons, listings, reviews, PR and external authority.
The important point is that the trend continued. Since 10 April, Odyssiant's visibility score has moved from 17 to 26, and on the comparable OpenAI prompt set it has moved from 17.4 to 27.2. That is not a claim of finished work. It is evidence of the method: audit, act, retest and improve.
Why we ran this on ourselves
The results were useful for two reasons. First, they showed where our own content was underperforming. Second, they showed what happened when we acted on that insight. As we made improvements to our content structure and content pillars, Odyssiant's scores increased across the middle and lower stages of the buyer journey — and we have kept auditing, acting and retesting since.
Between 10 April and 11 May, our overall visibility score moved from 17 to 23. More importantly, the improvement was not evenly spread. It happened exactly where we would expect better content architecture, clearer product messaging and stronger proof to have an impact.
A later comparison between the 28 April and 11 May audit runs showed 31 individual AI answers improved. That is the point of Odyssiant: not just to measure whether your brand appears in AI answers, but to show where buyer understanding is weak, what needs to change, and whether the work is making a measurable difference.
Buyer-journey stage movement
| Buyer journey stage | 10 April | 11 May | Change |
|---|---|---|---|
| Awareness | 0 | 0 | 0 |
| Consideration | 15 | 24 | +9 |
| Evaluation | 39 | 51 | +12 |
| Decision | 22 | 25 | +3 |
The problem: AI visibility is not one score
One of the mistakes in AI visibility reporting is treating it as a single number. That may be convenient, but it is rarely useful. A company can be invisible at the top of the journey and still appear in more specific product comparisons. Or it can be mentioned by name but misunderstood in the answer. Or it can appear in an answer but lose the recommendation to a competitor because the evidence base is stronger elsewhere.
That was exactly why we audited Odyssiant in the same way we audit customers. We wanted to understand:
- Whether Odyssiant appeared when buyers asked broad AI visibility questions.
- Whether the product was understood correctly.
- Whether AI answers reflected our actual differentiation.
- Whether our content pillars were strong enough to support buyer-stage questions.
- Which prompts were improving, declining or standing still over time.
The audit gave us a clear answer. Our lower-funnel explanation was stronger than our upper-funnel visibility. When AI systems had enough context to understand what Odyssiant did, the product performed increasingly well. But broader awareness was still weak. We did not just need more content. We needed better-structured content, clearer internal linking, stronger product-led pages, and more third-party authority over time.
What we changed
The main focus during this period was not a large backlink campaign. That is important. The improvements came primarily from work on content structure, content pillars and product explanation. We focused on making the Odyssiant proposition easier for AI systems to understand and retrieve. That included tightening the language around:
- Product AI visibility
- Buyer journey analysis
- Product-level scoring rather than brand-only monitoring
- Competitive visibility
- AI answer evidence
- Action recommendations across content, proof, PR, listings and third-party sources
- The difference between SEO/AEO tools and a marketer-first AI visibility platform
The aim was not simply to publish more. It was to make the existing content ecosystem more coherent. AI systems do not only look for keywords — they look for patterns of evidence. They need to understand what a company does, who it is for, how it is different, and why it should be included in an answer.
What happened
The clearest movement came in the parts of the journey where better content should have the greatest impact.
Consideration improved from 15 to 24
AI answers became better at recognising Odyssiant when buyers were exploring options, categories and approaches. That is meaningful because consideration-stage visibility is where buyers are still shaping their view of the market. They may not yet be asking for a specific vendor — they are asking what kind of solution exists, what the category means, how it compares to familiar alternatives, and which approaches are credible.
For Odyssiant this is a critical battleground. If AI systems understand the difference between technical SEO monitoring, brand-level AI visibility tracking and product-level buyer journey analysis, we have a much stronger chance of being represented accurately.
Evaluation improved from 39 to 51
This was the strongest stage movement. Evaluation is where buyers compare options, assess proof and test whether a product fits their problem. It is also where content quality matters most — better product pages, clearer positioning, stronger proof points and more structured explanations all help AI systems describe a product more accurately.
The +12 point movement suggests the content changes helped AI answers better understand how Odyssiant fits into the market and what makes it different.
Decision improved from 22 to 25
Decision-stage movement was smaller but still positive. That is not surprising. Decision-stage answers often depend on harder proof: reviews, comparisons, third-party validation, customer evidence, case studies, analyst-style references and wider authority signals.
The increase shows progress, but it also points to the next layer of work. To move decision-stage visibility further, Odyssiant needs more external proof, stronger off-site signals and more third-party evidence that AI systems can cite.
Awareness stayed at 0
This may look like the weakest part of the story. It is actually one of the most useful findings. Awareness did not move. That tells us the work improved how Odyssiant was understood in more specific buyer-journey contexts, but did not yet materially change broader category-level visibility.
That is exactly what we would expect at this stage. Awareness is often harder to shift with on-site content alone. It usually needs broader authority: backlinks, external mentions, third-party sources, media coverage, category pages, community discussion and citations from trusted domains. Different stages need different actions.
Why this matters
This is an ongoing case study, but the lesson is already clear. Across repeated audits, Odyssiant has shown measurable movement in the places where content improvements should have the biggest effect. The data also avoided a common trap: pretending everything improved. It did not. Awareness stayed flat. That gives us a better strategy, not a worse one.
Structured content improvements can change how AI systems understand and represent a product in consideration, evaluation and decision-stage answers. But broader awareness still needs external authority, citations and backlinks.
That is a more useful finding than “content fixes everything”. AI visibility is not random. It can be diagnosed. It can be improved. And different weaknesses require different actions.
What Odyssiant showed us
Product-level visibility matters more than brand visibility
A brand can be known and still fail to appear for the products or services that matter. The real question is not “Does AI mention us?” It is “Does AI understand what we sell, who it is for, when to recommend it, and why it is different?” Our improvement was strongest where the product story became clearer.
The buyer journey matters
A single visibility score would have hidden the pattern. Awareness did not improve. Consideration, Evaluation and Decision did. That distinction matters because the actions are different — weak Evaluation needs better product explanation, comparisons and proof; weak Awareness needs broader authority, PR and category-level content.
AI visibility needs action, not just monitoring
The most valuable part of the audit was not the score. It was the direction. The data showed what to fix, where to focus, and what kind of work was likely to move the result. Measurement is useful only if it leads to better marketing decisions.
What happens next
The next phase for Odyssiant is still external authority, but the 31 May audit gives us a more specific action plan.
The content pillar work has improved mid-to-late journey visibility. The latest run shows that Odyssiant is strongest in Evaluation and Decision, where buyers are comparing tools, assessing fit and looking for proof. The next opportunity is to improve broader awareness and strengthen consistency across engines, especially Claude. That means more work on:
- Backlinks
- Third-party citations
- Case studies
- Product comparisons
- Listings and directories
- Reviews and proof points
- PR and media coverage
- Category-level educational content
- Stronger internal linking around the core content pillars
- AI-readable company information and structured product evidence
- More external proof for claims around product-level AI visibility
The aim is to improve not just how Odyssiant is described when it appears, but how often it is surfaced in the first place. That is the difference between answer quality and answer inclusion. Both matter — but they do not move in the same way.
The takeaway
This was a useful test because it was real. We audited ourselves, found weaknesses, made changes and measured the movement. The result was not a miracle jump. It was something more useful: a credible improvement pattern.
In the first month, our own audit showed:
- Overall visibility increased from 17 to 23 between 10 April and 11 May.
- Consideration increased from 15 to 24.
- Evaluation increased from 39 to 51.
- Decision increased from 22 to 25.
- 31 individual AI answers improved between the later April and May audit runs.
- Awareness stayed at 0, showing that broader authority work was still needed.
The 31 May update added a broader test:
- Overall visibility reached 26 in a four-engine audit.
- The latest run tested 520 answers across OpenAI, Gemini, Perplexity and Claude.
- On the comparable OpenAI prompt set, average visibility increased from 17.4 to 27.2.
- 38 comparable prompts improved, 8 declined and 59 stayed the same.
- OpenAI and Perplexity citation coverage remained very high at 99%.
- Claude remained the weakest engine, showing where further work is needed.
- Awareness remained the hardest stage to move.
That is what eating your own dog food should mean. Not using your own product to create a polished marketing claim. Using it to find the uncomfortable gaps, take action, and prove whether the work made a difference over time.