By mid-2026, more than half of online searches that start with a question end inside an AI-generated answer instead of a traditional list of blue links. ChatGPT search, Perplexity, Claude's web tool, and Google's AI Overviews collectively serve hundreds of millions of synthesized answers per day. Each of those answers cites two to five sources. Getting cited is the new ranking goal.
The discipline of building websites so they get cited is called Generative Engine Optimization, or GEO. It overlaps with traditional SEO but optimizes for a different end state: instead of "page one of Google," the goal is "footnote source on a Perplexity answer." Most San Diego small businesses are not yet doing any of this work. The ones that start now will compound traffic that competitors can't catch up to.
The 30-second answer
To get cited by AI engines, a website needs five things: structured data (JSON-LD schema), an llms.txt file, semantic HTML, FAQ blocks written for verbatim quoting, and Core Web Vitals tuned to fast crawler indexing. Most of these are one-time structural fixes — they don't require ongoing content effort to maintain.
Why AI search citations matter (and the actual numbers)
Pages cited in Google AI Overviews earn 35% more organic clicks than non-cited competitors on the same SERP, according to multiple 2025 industry studies. Perplexity-sourced traffic converts at roughly 11x the rate of traditional organic traffic. ChatGPT's web search drives meaningful traffic for service businesses in major US metros, often surfacing local providers that aren't ranking in the top 3 of traditional Google.
The flip side: AI engines cite a much smaller pool of pages than traditional search. Google indexes hundreds of pages per query; an AI engine cites two to five. The selection criteria are tighter. Without the right structural fixes in place, a site never enters the citation pool regardless of how well-written its content is.
The five-fix priority order
Fix 1: Install JSON-LD schema on every page
JSON-LD schema is structured data that tells machines what a page is about. Google has supported it for over a decade for rich results; AI engines now use it as a primary signal for entity recognition.
For a San Diego small business, the minimum schema set is:
- LocalBusiness schema on the home page. Includes the business name, address (or service area), phone, hours, and category. This is what tells Google and AI engines "this is a real San Diego business with a physical service area."
- Person schema for the owner / founder. Connects the business to a real human entity. Cite alma maters, certifications, and bio facts.
- Service schema on every service page. One per service. Says "this page describes the X service offered by the LocalBusiness."
- FAQPage schema on pages with FAQ sections. Each Q&A becomes a citable answer that AI engines can quote verbatim.
- BreadcrumbList schema on every interior page. Helps AI engines understand the site's information architecture.
Schema must be in static HTML (not injected by JavaScript at runtime) for AI engines to reliably read it. Most schema injection apps inject via JS — that works for Google but is unreliable for AI engines that don't execute JS.
Fix 2: Publish a clean llms.txt file at the root
llms.txt is the AI-engine equivalent of robots.txt: a curated map of the site's most important content, structured for LLMs to consume. A typical llms.txt includes the brand entity description, a list of services with one-line summaries, key FAQs, and a positioning statement.
The file should live at https://[your-domain].com/llms.txt. On Webflow this requires either a custom Cloudflare Worker or a published page at /llms-txt with content rendered as markdown. On Shopify it's typically a custom asset upload. On Wix and Squarespace it's harder to ship cleanly — one of the structural advantages of building on Webflow.
The content of llms.txt should be 500-2000 words of curated brand summary. Not the full website — just the entry point map.
Fix 3: Write FAQs in answer-first format
AI engines love FAQ-formatted content because Q&A is the natural shape of LLM training data. Each FAQ block should follow the pattern:
- Question phrased as a search query
- Answer that opens with the direct answer in the first sentence
- Supporting context in subsequent sentences
- Concrete data points, named entities, and specific numbers wherever possible
The first sentence is what gets quoted. The rest is what gives the AI engine confidence the source is real. AI engines penalize answers that bury the lede with five paragraphs of intro before the answer.
Wrap each FAQ block in FAQPage schema. The combination of visible answer-first text + structured data is the single highest-leverage citation signal.
Fix 4: Use semantic HTML, not div soup
AI engines parse HTML structure to understand pages. A page built with proper H1 / H2 / H3 hierarchy, <article> tags, <section> tags, and <nav> landmarks is dramatically easier to parse than a page built with nested div containers and class-based styling alone.
This is one of the structural reasons Webflow ranks above Wix and Squarespace for AI search readiness: Webflow's output is significantly cleaner HTML5 semantic markup. WordPress varies by theme. Custom-coded sites can be excellent or terrible depending on the developer.
For a San Diego small business choosing a platform with AI search in mind: Webflow is the cleanest, Shopify is acceptable, Squarespace is constrained, Wix is most constrained.
Fix 5: Tune Core Web Vitals to green
AI-engine crawlers (perplexity-bot, GPTBot, Claude-Web, Google-Extended) have time budgets per crawl. Slow pages get fewer pages indexed per crawl session, which means new content takes longer to enter the citation pool. Pages with poor Largest Contentful Paint (LCP) or Cumulative Layout Shift (CLS) can be skipped entirely.
The targets:
- LCP under 2.5 seconds on mobile
- CLS under 0.1
- Total Blocking Time under 200ms
Common San Diego small business sites fail these because they load 10+ Google Font families, run multiple analytics scripts in the head, and serve unoptimized hero images. The fix list: trim font families to 2-3 max, defer all non-critical JS, serve images as WebP with srcset.
Off-site signals that compound the on-site work
The five fixes above are on-site. They get a site into the citation pool. Off-site signals determine which sites in the pool get cited most.
For San Diego small businesses, the off-site work that matters:
- Google Business Profile. Verified GBP listings get cited by AI engines for local-intent queries at much higher rates than non-verified entities.
- Directory citations. Yelp, Bing Places, BBB, industry-specific directories. Each citation reinforces the entity graph.
- Social profile sameAs links. The LocalBusiness schema's sameAs array should link to every social profile. This tells AI engines "this brand and these profiles are the same entity."
- Backlinks from real publications. A single editorial mention from a respected industry publication is worth more than 100 directory citations.
How long until results show up
Structural readiness is delivered within the first 30-60 days of a GEO retainer. Actual citation appearances typically start showing up within 60-120 days as AI engines crawl the new schema and index the FAQ content.
The signal to watch is in Google Search Console's Performance reports: when a site starts showing up for queries it's never ranked for before, that's usually because an AI Overview cited it and the click flowed back to GSC. Perplexity's referral traffic shows up in GA4 under unassigned channel traffic and can be identified by user-agent strings if you have advanced analytics.
What it costs to do this work
The structural fixes (#1-5 above) are typically a one-time GEO Implementation Sprint — a 2-4 week engagement that installs all the schema, writes the llms.txt, audits Core Web Vitals, and rewrites the FAQ blocks across the site. After that, ongoing AI search optimization is mostly a content cadence question: writing pillar articles, keeping the schema fresh, and monitoring citation appearances.
For brands that want ongoing optimization beyond the sprint, an AI + SEO Retainer covers monthly content, schema updates, GBP management, and citation monitoring. Most San Diego small businesses see meaningful traffic shifts within 90-120 days of a retainer.
How Kanyon Studio approaches GEO
Every AI search engagement starts with an audit: what's on the site, what's missing, what's the citation potential, what's the off-site visibility. The audit is a deliverable in itself — you can decide whether to engage further or hand the audit to a different vendor.
From there: GEO Implementation Sprint to install all structural fixes, then optional retainer for ongoing optimization.
See the AI Search & SEO service page → · Book a discovery call →
