Restructuring CarWale's information architecture and on-page semantics to reverse a multi-year decline in organic search visibility, with a particular focus on the car research and comparison pages that drive high-intent commercial traffic.
CarWale sits on one of the deepest content libraries in the Indian automotive web, covering thousands of car variants, model reviews, comparison pages, and dealer locations. The site had ranked well historically and continued to receive substantial organic traffic, but the underlying picture was not as healthy as the topline number suggested.
Organic visibility was drifting downward over consecutive quarters, and the pattern was specific. Pages were ranking, but they were ranking for shoulder queries rather than high-intent commercial terms. A user searching for general information about a model would find us. A user searching for the kind of query that signals real buying intent (model versus model comparisons, variant-specific queries, on-road pricing in specific cities) was increasingly finding us further down the results page than we should have been.
This is the worst kind of SEO problem because the topline traffic numbers hide it. Sessions look stable while the commercial value of those sessions silently erodes. Internal stakeholders saw flat traffic and assumed everything was fine. The SEO team saw the underlying intent shift and was raising concerns that were not getting prioritised. The brief I took on was to translate the SEO team's diagnostic into design and product changes that the broader org could rally around.
I partnered with the SEO lead to run a diagnostic across roughly 200 high-priority page templates. The exercise had to be structural rather than page-by-page, because we could not redesign a thousand individual pages and the template was where the leverage lived anyway.
Three structural issues surfaced. Cannibalisation: similar pages were competing with each other for the same intent. A query that should have anchored on a single canonical page was instead spread across three or four near-duplicates, each diluting the signal of the others. Inconsistent semantics: our HTML structure varied across templates, which made it harder for search engines to reliably extract entity relationships from our pages. Different templates used different heading hierarchies, different schema markup, different patterns for the same information. Thin content on high-intent pages: comparison and decision-making queries needed substantive content to rank, and several of our highest-commercial-value templates had content depth that was below the threshold a query like 'Hyundai Creta vs Kia Seltos' actually requires.
The diagnostic became a template-level brief rather than a page-level audit. Each template got a one-page brief that translated SEO intent into design requirements: which entities had to be present, what schema markup was required, what content depth was appropriate, and what the right primary calls to action were. Designers stopped having to guess at SEO implications and SEO stopped having to argue from the sidelines.
The framing shift that unlocked the rest of the work was deciding to treat each URL as a product surface with explicit users, goals, and success criteria, rather than as a passive output of the content team. This sounds like a small shift and it changed everything.
For each high-priority template, the redesign followed a structured approach. The team would start from the queries the template was meant to serve, then map those queries to the entities and relationships search engines were looking for, then design the page hierarchy to surface those entities in the order that matched the user's decision process. The page hierarchy was downstream of the query intent, not upstream of it.
This is the part most SEO-conscious redesigns get wrong. Teams tend to start from the existing page, identify SEO issues, and fix them in place. That preserves the original page's structure and patches around it, which produces incremental improvement and locks in the structural problems. Starting from the queries forces a clean structural answer.
In parallel, I worked with engineering to standardise the semantic HTML patterns across templates. Heading hierarchies, schema markup, microdata, and structured data all converged on a single shared vocabulary that every template implemented identically. Search engines reward consistency more than they reward optimisation, and the consistency win was as important as any individual page change.
Consolidate cannibalising URLs before redesigning anything. The instinct on a project like this is to redesign every weak page. The smarter early move was to first consolidate the URLs that were competing with each other. Several pages were inadvertently splitting the signal for the same query, and the fix was a redirect strategy rather than a redesign. Once consolidated, the surviving pages had a much higher signal for search engines to read, which made the subsequent redesign work pay off faster than it otherwise would have.
Move content into the rendered DOM, even if it lives behind tabs visually. Many of our pages had content hidden in tabs or accordions that loaded only on user interaction. Search engines vary in how reliably they extract that content. We moved the content into the default rendered DOM and let CSS and JavaScript control what was visually shown, which kept the user experience identical while ensuring search engines saw everything.
Schema markup as a first-class design deliverable. Schema is usually an afterthought that engineering adds after the design is done. I treated it as a design decision, with the schema requirements documented as part of the template brief and the schema fields visible in the design specification. This made schema implementation faster and more accurate, because designers were thinking about it from the start rather than retrofitting it.
A brief-driven workflow that survived the rollout. Subsequent template work now starts with an SEO-and-user brief rather than design choosing one and SEO finding out later. The collaboration model itself became the durable outcome.
Information density without visual clutter
Organic visibility recovered on high-intent commercial queries over the quarters following the rollout, with the comparison and variant pages benefiting most. The shape of the improvement was characteristic of structural fixes: a gradual lift over multiple weeks as search engines re-crawled and re-evaluated the templates, rather than a sudden spike.
Just as important as the visibility recovery was the internal process shift. The brief-driven workflow stuck. Designers now treat SEO not as a constraint to comply with but as a research signal to design from. The queries people type into Google are some of the cleanest user-intent data available, and the team learned to read them that way.
The most lasting outcome was the template library itself as a system. New templates that the team has built since the original work follow the same patterns and inherit the same structural discipline, which has compounded the original effort across pages we did not directly touch.
The biggest internal shift was getting designers to stop seeing SEO as a separate department's compliance checklist and start seeing it as a research signal. Search query data is some of the most honest user-intent data available, and treating it as a brief input rather than a downstream optimisation changed the texture of the work the design team produced.
I would push that mindset earlier next time so the team does not lose months treating SEO as someone else's problem. The work would also have moved faster if I had spent more time upfront aligning the content team on the new template patterns, because content production lagged design by a quarter and held up the visibility recovery on several templates.