Published with permission from AgaOne Commodities.
Key findings
- 1
Both sites launched with considered design, working UX and standard SEO in place. Both scored in the mid-fifties on AI readiness.
- 2
The gaps were not content quality. They were crawler access, structured data, and the identity signals a machine uses to decide whether a source is credible.
- 3
The fixes were technical and took days, not months. One site reached 100 out of 100 on the checks Citehound runs today. The other reached 89.
01
The situation
AgaOne Commodities launched two new corporate sites. Planning, content and launch were handled in house, with traditional SEO applied throughout.
The marketing manager who ran the project scanned both sites after launch. Not because anything looked broken, but because the question had changed. Buyers were finding suppliers through answer engines as well as through Google, and neither site had been built with that in mind.
“AI visibility should begin during development, not after launch.”
— Saruhan Efe Saruhanoğlu, Marketing Manager, AgaOne Commodities
Both sites came back in the mid-fifties. Site one scored 58. Site two scored 55.
02
What the scans found
The failures clustered in two of the three pillars, and neither was about how the sites read to a person.
Discoverability. robots.txt and sitemap structures were not set up for AI crawler access. Neither site declared an llms.txt.
Content and trust. Organization and WebSite schema were missing. Heading hierarchy was inconsistent. The company’s identity, markets and contact details were legible to a reader and ambiguous to a machine.
The technical foundation pillar was closer to complete, which is the pattern we see across most sites that were built with SEO in mind. Canonical tags, language attributes and titles are familiar work. Schema and crawler policy are not.
03
What changed
The work ran in the order the report ranked it, highest impact first.
Crawler access first. robots.txt and sitemap structures were reviewed so that important pages could be discovered. AI crawler permissions were checked and corrected. Nothing further matters if a crawler cannot reach the page.
Then machine readability. Canonical tags, language declarations, titles, meta descriptions, Open Graph tags and JSON-LD were implemented or corrected, depending on what each site was missing.
Then identity. Organization and WebSite schema were added. Heading hierarchies were restructured. Company information, services and markets were presented so that the relationship between the brand and its published information was explicit rather than implied.
Both sites were then scanned again, and the remaining gaps were worked through in a second pass.
“A measurable score and a structured checklist made communication between marketing and technical teams more precise. It also made tasks easier to prioritise, implement, and track.”
— Saruhan Efe Saruhanoğlu, Marketing Manager, AgaOne Commodities
That second sentence describes something we did not design for and have seen repeatedly since. The score is useful to a marketer mainly because it converts into a list a developer can act on.
04
Results
| Initial | Final | Change | |
|---|---|---|---|
| Site one | 58/100 | 100/100 | +42 |
| Site two | 55/100 | 89/100 | +34 |
| Average | 56.5 | 94.5 | +38 |
Fig. 1
Source: Citehound scans, August 2026, initial and final, n=2 sites
Site two’s remaining gaps were recorded as a roadmap rather than closed, which is the right outcome. Some of what Citehound flags takes a content decision rather than a code change.
“Thanks to the action plan provided by Citehound, we increased the AI visibility score of our two websites by an average of 38 points.”
— Saruhan Efe Saruhanoğlu, Marketing Manager, AgaOne Commodities
05
On the perfect score
One site reached 100. That number needs explaining, because it means something narrower than it looks.
A score of 100 means the site passes all 16 checks Citehound runs today. It does not mean the work is finished, and it does not mean the site will be cited in an AI answer. It means every signal we currently measure is present and correct.
Our scoring model is going to get harder. We are adding checks for agents.md, product-page schema and commerce discovery endpoints, and we are building a full-site crawl that scores every page rather than the homepage alone. A site that scores 100 on its homepage today will not score 100 across fifty pages, and we expect these two sites to move when the new checks land.
That is the intended behaviour. A readiness score is a measure of the current standard, and the standard is moving. We will re-scan both sites as the model expands and publish what changes.
What this does not tell you
Citehound measures AI readiness. It reads what a site publishes — robots.txt, sitemap.xml, llms.txt and the page itself — and scores whether AI systems can reach it, parse it, and find the signals they use to weigh a source.
It does not measure how often ChatGPT, Claude or Perplexity name a brand today. Those are different questions, and a score of 100 is not a guarantee of citation.
The reverse holds more firmly. If a crawler cannot reach a site, or cannot interpret what it finds, the chance of that content being evaluated as a source drops sharply. Readiness is a precondition, not a promise.
Scanned: August 2026 · Checks: 16 across three pillars · Method: methodology
How to cite this
Andaç Üzel, “Two Sites, Launched With Good SEO, Invisible to AI Crawlers,” Citehound Case Studies, August 2026. https://getcitehound.com/research/case-study-agaone
More case studies will follow as clients agree to share their numbers. If you want to see where your own site lands, run the free scan.
Scan your site free