What our client got
| Business metric | Before | After | Change | Change % |
|---|---|---|---|---|
| Average weekly organic traffic | 27,495 | 32,394 | +4,899 | +18% |
| Average weekly non-branded organic traffic | 3,219 | 7,645 | +4,426 | +138% |
| Keywords in positions 1-3 | 142 | 303 | +161 | +113% |
| Keywords in positions 4-10 | 393 | 443 | +50 | +13% |
| Total tracked organic keywords | 727 | 910 | +183 | +25% |
| Weekly positions in the top 10 | 1,548 | 2,532 | +984 | +64% |
| AI Overview citations | 106 | 306 | +200 | +189% |
| Share of Voice | 10.4% | 20.6% | +10.2 pp | +98% |
Source: Ahrefs. Comparison period: May 1 to July 1, 2026.
Before the rendering fix, AI systems could not retrieve the full source content and hallucinated facts and product details. After the release, they could read and cite the product information the platform had published.
Why were the products missing from AI search?
Search and AI systems could not read the content
JavaScript hid meaningful body content across thousands of pages on the platform. Search and AI systems could reach the URLs, but could not reliably retrieve or interpret the published product information.
Crawlers could not reach all intended pages
Sitemaps omitted indexable URLs. Broken internal links ended at dead pages, while redirects added unnecessary steps between crawlers and useful content.
Shared site rules repeated the same errors
Metadata and structured-data rules produced defects across whole page groups. Fixing the shared rules allowed each release to correct thousands of pages at once.
We ranked the audit findings by impact, turned them into release-ready tickets, and repeated the same checks after every deployment.
How we moved the strategy into production
- 1
We agreed on success with the CMO
The CMO set the business goal: grow organic traffic. We agreed on the technical measures that would track progress, reported them during the engagement, and measured the search results at the end. - 2
We found the shared constraints
We used Dioptria, our custom audit-grade crawler system, with Search Console data and page-level checks. We traced visibility problems to the routes, layouts, and components causing them, then ranked the work by affected scope and expected technical effect. - 3
We gave engineering release-ready work
Every priority identified the affected system, the required behavior, the business reason, the acceptance check, and the post-release test.Every Asana ticket carried
- the affected route, layout, or component and the source of the problem
- the required behavior, business reason, and priority
- the exact difference, acceptance check, and post-release test
- 4
We worked through every release
Every week, BeRelevant.ai's team, our client's product manager, developers, and internal SEO specialist reviewed progress, answered implementation questions, and agreed on the next work. We met with the CMO every two weeks to report against the agreed measures. Our client's developers changed and released the code. We crawled the site again after every release, compared the result with the acceptance check, and either closed the issue or returned it with fresh evidence.
What we changed
Dioptria audit
| Technical result | Discovered | Remaining after fixes | Change | Why it mattered |
|---|---|---|---|---|
| Pages withholding body content behind JavaScript | ~8,000 | 0 | -8,000 (-100%) | AI systems could read and cite the published product information |
| Indexable pages missing from sitemaps | 4,266 | 12 | -4,254 (-99.7%) | Crawlers could discover the intended pages |
| Broken internal page targets | 1,176 | 4 | -1,172 (-99.7%) | Crawlers stopped hitting dead ends |
| Internal links passing through redirects | 547 | 50 | -497 (-91%) | Crawlers gained direct paths to content |
| Metadata field defects | 41,557 | 969 | -40,588 (-98%) | Search systems received cleaner page-level signals |
| Reused metadata instances | 24,332 | 2,371 | -21,961 (-90%) | More pages carried distinct titles and descriptions |
| Conflicting metadata values | 16,434 | 7,790 | -8,644 (-53%) | Search systems received fewer contradictory signals |
| Pages missing their intended structured-data type | 2,352 | 0 | -2,352 (-100%) | Search and AI systems received explicit page meaning |
| Hreflang links pointing to error pages | 874 | 0 | -874 (-100%) | Crawlers reached the correct language destinations |
Discovered = highest count across weekly audits from May 7 to July 5. Remaining = July 5 audit.
The order mattered. Rendering exposed the full content first, which allowed our crawler to find deeper link, sitemap, metadata, and schema defects. We then traced those defects to shared templates and rules, so each engineering release could fix thousands of pages at once.
The weekly crawl audit measured what the site served after each release. We used it to accept the fix or return it to development. We used Search Console to measure how indexation and field performance changed across the engagement.
Google Search Console
| Technical result | Before | After | Change | Why it mattered |
|---|---|---|---|---|
| Pages indexed by Google | 6,014 | 7,491 | +1,477 (+25%) | More pages became eligible to rank and surface in search |
| Desktop URLs rated “poor” for Core Web Vitals | 296 | 1 | -295 (-99.7%) | Google cleared the poor classification from nearly every affected URL |
| Desktop URLs rated “needs improvement” | 674 | 24 | -650 (-96%) | More pages passed stronger field-performance thresholds |
Source: Google Search Console. Comparison: May 1-2 to June 30-July 1, 2026.
At the end of the engagement, we used Search Console to measure the broader technical results and Ahrefs to measure the change in search and AI visibility. The outcome for the business: customers can now discover the platform’s products through non-branded search and AI answers, where the products were previously invisible or misrepresented.
We carried canonical, robots-meta, mobile performance, external-link, and smaller metadata issues into the next phase.
How we would start our engagement
Set the commercial priorities
We begin with the products that generate revenue today, the products you want to grow, the markets that matter, and what you want customers to find through search and AI.
Decide what needs to become visible
We determine which content needs to be created, improved, or exposed to search and AI systems. Then we connect those needs to the site changes required to earn visibility.
Take the work through release
We order the work around your commercial priorities, take the plan through marketing, product, and engineering, verify each release, and measure the results against the business goals agreed at the start.