Only 4.1% of AI citations point back to a brand's own website. The other 95.9% live on Amazon, Reddit, YouTube, editorial sites, and marketplaces. So on-site content work looks like a small lever. It is a small lever. It is also the lever that decides what happens on the citations that DO land on your domain, and it is the lever that decides whether AI engines can extract clean facts when they arrive. Across the 80+ Indian brand audits Cited has run in the past quarter, the same five on-site mistakes show up more than any others. Each strips the on-site surface of the signals AI engines use to pick you. Each is fixable in an afternoon.
This is the practitioner playbook, not a theory piece. The five mistakes come out of the June 2026 Cited Index dataset (226 Indian brands, 10 categories, 760 citations across ChatGPT, Perplexity, Gemini, Google AIO, and AI Mode), the GEO Score signal-level scans that back it, and audit patterns Salman has walked through with founders this quarter. Each mistake gets the failure mode, the fix with a code snippet, and a browser DevTools check to verify the fix landed.
Mistake 1: No author attribution
The biggest single delta between top-tier and mid-tier brands in the June 2026 Cited Index dataset is Author Attribution. 27 of 70 mid-tier Indian brands (ranks 4-10 in their category) fail this signal. Almost 2x the next-most-common technical weakness (Sitemap Accessibility, flagged for 15 mid-tier brands). Travel & Luggage is the cleanest case: 6 of 7 mid-tier brands (Briggs & Riley, Mokobara, Travelpro, Uppercase, Nasher Miles, Skybags, Rimowa) fail; Samsonite, American Tourister, Safari all pass.
AI engines inherit the E-E-A-T evaluation path (Experience, Expertise, Authoritativeness, Trustworthiness) from Google's Search Quality Rater Guidelines, because their training data already encoded authorship signals as quality markers. A page attributed to "Editorial Team" or to nothing at all gets de-weighted against a page with a named human author linked to a bio. ChatGPT reads JSON-LD author metadata when it constructs E-E-A-T scores. Perplexity surfaces author attribution in its citation panel when the data is there. Gemini and AI Mode inherit Google Search's higher trust bar for Your Money Your Life categories, which spans Health & Wellness, HR & Payroll, and CRM.
The fix. Add a JSON-LD Article block to every blog post and article page in your template head:
<script type="application/ld+json">
{
"@context": "https://schema.org",
"@type": "Article",
"headline": "Your article title",
"datePublished": "2026-07-13",
"dateModified": "2026-07-13",
"author": {
"@type": "Person",
"name": "Author full name",
"url": "https://yourbrand.com/authors/author-slug"
}
}
</script>
Then render a visible byline below the article title with rel="author" linking to the bio page. Build the bio page: name, role, headshot, 100-300 words, links to LinkedIn and prior published work. Full playbook: Author Attribution: The Universal Mid-Tier GEO Weakness.
Verify. In DevTools console on any article page: [...document.querySelectorAll('script[type="application/ld+json"]')].flatMap(s => { try { return [JSON.parse(s.textContent)] } catch { return [] } }).map(d => d.author).filter(Boolean). If the output shows your author object, the schema-backed signal is live.
Mistake 2: Thin entity content on brand and product pages
The second failure mode is pages that read as marketing prose instead of extractable entity content. AI engines pull facts as (entity, attribute, value) triples. A hero paragraph that says "We're on a mission to bring you the best wellness" gives an extractor nothing. A hero paragraph that says "Pure Nutrition is a Mumbai-based wellness brand founded in 2016, selling 60+ SKUs across nutraceuticals, immunity boosters, and skincare, priced ₹299-₹2,499, sold at purenutrition.me and Amazon" gives an extractor five facts to lift.
This mistake shows up hardest in categories where the brand's own site is the dominant citation source. In the June 2026 Cited Index, brand-site share of citations is 78% in Conversational AI, 71% in HR & Payroll, 69% in CRM & Sales, and 42% in Travel & Luggage. In those categories, the engine reaches your homepage and reads what it finds. If what it finds is a hero video and three "our story" paragraphs, the citation slot goes to whoever wrote extractable copy.
The fix. Rewrite the hero + spec sections on your product and category pages as entity-with-attributes structure. Add a Product schema block:
{
"@context": "https://schema.org",
"@type": "Product",
"name": "Product name",
"brand": { "@type": "Brand", "name": "Your brand" },
"category": "Category",
"description": "One-sentence factual description with concrete attributes",
"offers": { "@type": "Offer", "price": "1499", "priceCurrency": "INR" }
}
Verify. Paste the page URL into Google's Rich Results Test. If the Product entity renders with your name, brand, description, and offer, an AI extractor can pull the same triples.
Mistake 3: Missing FAQPage schema on Q&A content
Want to know how your brand scores on these same metrics?
We'll run 20 prompts across 3 AI platforms and send your report within 24 hours.
Q&A blocks on your pricing page, your how-it-works page, and your category buyer's guides answer prompts that AI engines are running against you every day. If those blocks carry FAQPage schema, engines can identify the content as answer-friendly and lift it into responses. If not, the engine treats it as ordinary body copy and often skips it.
FAQPage schema is the single most under-used citability lever across the mid-tier Indian D2C brands in Cited's audit sample. It maps to Cited 8 Metric 7 (Schema & Technical Health) and it sits at Layer 2 (Citability) in the 3-Layer AI Visibility Stack. Google's May 2026 guide says structured data is not required for Google AI Overviews specifically, and that is accurate for Google. It matters for ChatGPT retrieval, for Perplexity, and for Google AI Mode. The five-platform reality is that you ship it and let each engine use what it uses.
The fix. On any page with a real Q&A section, add:
{
"@context": "https://schema.org",
"@type": "FAQPage",
"mainEntity": [
{
"@type": "Question",
"name": "Your actual question here",
"acceptedAnswer": {
"@type": "Answer",
"text": "Your actual answer, 40-120 words, plain prose."
}
}
]
}
Only mark up questions that appear as real Q&A on the visible page. Google penalises FAQPage schema on content that is not truly Q&A.
Verify. Rich Results Test again. Confirm each Question and Answer pair renders.
Mistake 4: Tag proliferation without pillar hierarchy
Most brand blogs Salman audits have 500 auto-generated tag pages and five real pillar pages. The tag pages are thin, near-duplicate, and outrank the pillar pages on internal link equity. AI engines crawl the site, see 500 tag-page URLs and 200 blog-post URLs pointing at those tag pages, and lose the shape of what the brand is actually about. The result is that when an engine samples the site for grounding, it lands on a tag page called /tag/skincare-tips instead of the pillar page /skincare-guide, and the extracted context is generic.
This is a Layer 1 (Discoverability) failure that costs Layer 2 (Citability). The engine can find your content. It just cannot tell what content is central. Cited 8 Metric 7 (Schema & Technical Health) covers the schema half of this; the pillar-hierarchy half sits inside the same metric.
The fix. Audit your tag URLs once and keep the tags that map to 5-10 real content pillars. For every remaining low-value tag URL, pick one of two dispositions:
Aggressive (recommended for a big cleanup). 301-redirect the low-value tag URL into the pillar page that best fits its theme. The tag URL stops existing, and its link equity flows into the pillar page. Cleanest for AI-engine grounding because engines stop landing on thin pages entirely.
Conservative (recommended when inbound links are unclear). noindex, follow the tag page and consolidate internal links to point at the matching pillar page. The tag URL still exists but drops out of the AI-engine grounding pool.
For the noindex approach, add via meta tags:
<meta name="robots" content="noindex, follow">
on every low-value tag URL. Rebuild sitemap.xml to exclude noindexed URLs and ship self-referencing canonical tags on pillar pages.
Verify. Google Search Console → Pages report. Watch the "Not indexed" count for tag URLs move up and the "Indexed" count stay stable on your real content URLs.
Mistake 5: No content freshness signals
Perplexity's published retrieval research shows content under 1 year old is 3x more likely to be cited than content over 2 years old. Cited 8 Metric 8 (Content Freshness Rate) is the metric that measures this and it sits at Layer 1 (Discoverability). Two signals matter: a dateModified field in Article schema (schema-backed) and a visible "Updated" line above the fold (DOM-visible).
Most brand blogs Salman has audited this quarter carry the datePublished field only. When the brand refreshes an article in 2026, the schema still shows a 2024 timestamp. AI engines de-rank the page as stale, even though the content is current. The fix is one line of Liquid or Handlebars or JSX per template.
The fix. In the Article JSON-LD block (already added in Mistake 1), keep dateModified current:
{
"datePublished": "2024-08-14",
"dateModified": "2026-07-13"
}
Above the article title in the template:
<p class="article-meta">Updated 13 July 2026</p>
For evergreen content, add a lightweight changelog at the bottom of the article naming what changed on each refresh. This gives an AI extractor evidence the page is actively maintained, not just re-stamped.
Verify. Rich Results Test → confirm dateModified is present in the Article schema. Reload the page → confirm the visible "Updated" line renders.
Closing the loop
The five mistakes above are not the reason most Indian brands are under-cited by AI engines. The reason most Indian brands are under-cited by AI engines is off-site absence: they do not show up on the Reddit threads, marketplace listings, YouTube reviews, and editorial roundups that carry 95.9% of AI citations. That is the bigger fix and it takes months.
The five on-site mistakes are what strip the remaining 4% of your own citation opportunity. And they compound. A brand with named authors, extractable entity content, FAQPage-tagged Q&As, a clean pillar hierarchy, and current freshness signals is a brand AI engines can cite cleanly when they do land on your site. A brand missing all five is a brand engines have five reasons to skip.
Run a free GEO Score scan on your domain today to see which of the five you are missing. Ship the fixes this week.