By the end of 2025, 30 to 50 percent of top-funnel research queries ended in an AI Overview, a ChatGPT browse answer, a Gemini summary, or a Perplexity citation. Traditional SEO still matters, but the buyer journey now branches: they ask the assistant first, click the second-best source second. Below is the operating model Dcrayons uses to get clients cited across the four AI surfaces that count.
The four citation surfaces that matter in 2026
Not all AI surfaces are equal. Each pulls from a different signal mix:
- Google AI Overviews: pulls from top-ranking SERPs + structured data + entity graph. Heaviest overlap with classical SEO.
- ChatGPT browse: pulls from web index + OpenAI's preferred-citation list. Weighted toward authority domains and recent freshness.
- Gemini: pulls from Google's index + Gemini-specific quality signals. Overlaps Overviews but applies stricter trust scoring.
- Perplexity: pulls from its own crawl + preferred sources. Most aggressive about citing 5 to 8 sources per answer; highest opportunity for niche brands.
The work to win each surface is different. Treating them as one channel is the most common mistake we see.
The Dcrayons share-of-answer audit
Before any optimisation, we measure where you currently sit. The audit covers:
- Top 200 buyer queries in your category, scored against all four surfaces.
- Citation count per surface per query (are you in the top 5 sources, top 10, or invisible?).
- Competitor share-of-answer for the same query set.
- Entity graph integrity check across Wikipedia, Wikidata, Crunchbase, G2, Clutch, and category-specific directories.
- Crawl trace for each surface's preferred bot (Googlebot, OAI-SearchBot, PerplexityBot).
The output is a single Score from 0 to 100 plus a prioritised remediation roadmap.
What moves the needle on each surface
Google AI Overviews
The signals that lift AI Overview inclusion most reliably:
- FAQ + HowTo schema, well-formed and matching the query phrasing.
- Author entity coherence: every page byline resolves to a Person schema with hasCredential.
- Internal-link clustering that mirrors the entity graph (topics, not just keywords).
- E-E-A-T pillars hardened: lastReviewed + reviewedBy on YMYL pages, source citations on stat claims.
ChatGPT browse
OpenAI's browse model leans on a curated source list. Getting added:
- Show up on the top 3 SERPs for the query (classical ranking still gates entry).
- Maintain a clean robots.txt that allows OAI-SearchBot.
- Ship structured llms.txt + a markdown mirror at /llms-full.txt.
- Earn at least 5 high-authority backlinks in the last 90 days.
Perplexity
Perplexity loves freshness and source diversity. The unlock:
- Publish at a weekly cadence in your category (their crawler scores recency).
- Include numeric claims with explicit sources -- Perplexity quotes the data point.
- Avoid paywalled content -- their crawler skips it.
The 90-day remediation plan
Once the audit lands, the typical 90-day plan looks like this:
- Weeks 1 to 3: Entity graph repair (Wikipedia, Wikidata, Crunchbase, G2, Clutch). Schema audit + fix. llms.txt + llms-full.txt ship.
- Weeks 4 to 7: Content engineering on the 30 highest-opportunity queries. FAQ + HowTo schema where applicable.
- Weeks 8 to 10: Authority signal work -- earned mentions on the source domains your category's AI surfaces cite most.
- Weeks 11 to 12: Re-audit + Score readout + next 90-day plan tied to one revenue metric.
The CFO-readable metric
Vanity metrics like "AI mentions" do not survive a CFO review. What we report monthly:
- Share-of-answer on the buyer's top 50 queries (weighted by intent).
- AI-attributed sessions (UTM-tagged where possible, GA4 source-filtered where not).
- AI-attributed pipeline (revenue tracked through the engagement window).
Where most brands fail
Three patterns we see in audits:
- Treating AI search as "SEO with extra steps." The two channels overlap but the optimisations diverge after week one. You need both teams aligned, not one team doing both.
- Skipping the entity graph. AI surfaces resolve brand identity through Wikipedia + Wikidata + Crunchbase. If those are wrong or thin, every other signal underperforms.
- No measurement window. AI-attributed revenue is murky. Without a deliberate measurement design, the program looks like it failed even when share-of-answer is moving.
Get the audit
Every Dcrayons engagement starts with the share-of-answer audit. It is free on the proposal call. The first month of work is fully refundable if the Score readout does not land in one business day.
