How to Rank in Google AI Overviews and Perplexity (2026 Guide)

The 9-step framework

This is the framework we use at Frostbite Marketing to get our clients cited by generative AI. It works for both Google AI Overviews (which lean on classic search rankings) and the open-web AI engines (Perplexity, ChatGPT, Claude, Bing Copilot, Apple Intelligence).

Step 1 — Win the underlying classic ranking first

Google AI Overviews are built from pages already ranking on the first page of Google for the query. Perplexity and Bing Copilot lean heavily on Google/Bing indexes via retrieval-augmented generation (RAG). If you’re not ranking in the top 10 for “best CRM for small business”, you’re not getting cited in any AI answer for that query either.

Implication: classic SEO is the foundation. There’s no AI-ranking hack that skips it. The good news — most of the techniques in steps 2–9 also improve classic rankings.

Step 2 — Answer the question in the first 50 words

AI engines extract answer paragraphs from your content. The first 40–60 words of the article should contain the literal answer to the question implied by the title.

Bad opener: “Customer relationship management is a fascinating discipline that has evolved significantly over the past decade. In this article, we’ll explore…”

Good opener: “The best CRM for most small businesses in 2026 is HubSpot Starter ($20/user/month) for sales-led teams, GoHighLevel for agency-built workflows, or Frostbite’s all-in-one CRM for businesses that want CRM + AI receptionist + marketing automation in one platform. Here’s how to choose between them.”

The second one answers the question, provides specific brands, mentions prices, and immediately tells the reader (and the AI) what the article will cover.

Step 3 — Use question-first H2s

H2s should mirror the questions real people ask. Treat the People Also Ask box for your target query as your H2 outline.

Example for “how to rank in google ai overviews”:

• H2: What are Google AI Overviews?
• H2: How does Google decide which pages to cite?
• H2: Do I need to be in the top 10 to be cited?
• H2: What schema markup helps with AI Overviews?
• H2: How long does it take to start appearing?
• H2: Can AI Overviews hurt my organic traffic?

Each H2 followed by a 100–200 word direct answer, then optional supporting detail.

Step 4 — Add the schema trio: Article + FAQPage + Person

Every blog post and resource page should output:

• Article (or BlogPosting) schema with headline, author, datePublished, dateModified, mainEntityOfPage
• FAQPage schema for any Q&A section
• Person schema for the author (linked from the Article schema)

AIOSEO Pro generates the first two automatically. The Person schema typically requires an author profile page on your site with credentials, photo, and sameAs links to social profiles. See our /llms.txt for the structured representation we publish.

Step 5 — Surface freshness signals

LLMs and AI engines actively favor recent content. The signals that matter:

• “Last updated: YYYY-MM-DD” visible in the article body (not just metadata)
• dateModified in the Article schema, updated when the content is actually refreshed
• A predictable annual update cadence on cornerstone pages
• Date references inside the content itself: “As of 2026…”

Bury the date and the engine assumes stale.

Step 6 — Build author E-E-A-T

The author is becoming a first-class ranking signal. Required artifacts:

• Author profile page at /authors/{slug}/ with bio, photo, credentials, contact, and social sameAs links
• Bylines on every piece of content linking to the author page
• Person schema on author pages
• External authority signals: podcast appearances, conference talks, contributed posts, press mentions, awards

For agencies and consultancies, the people doing the work need a public identity. Anonymous content is increasingly invisible.

Step 7 — Publish a comprehensive llms.txt

• Place it at /llms.txt
• Include: 2-3 sentence Organization summary, list of core services with links, list of locations with links, top blog posts with descriptions, contact info, citation preference statement
• Optionally publish a longer /llms-full.txt with comprehensive Q&A, FAQ-style facts, comparison data, and case studies

See /llms.txt for the structure we use.

Step 8 — Allow all major AI crawlers

Your robots.txt (and optional /ai.txt) should explicitly allow:

• GPTBot (OpenAI training)
• ChatGPT-User (ChatGPT browsing)
• OAI-SearchBot (ChatGPT search)
• ClaudeBot, Claude-Web, Claude-SearchBot, anthropic-ai
• PerplexityBot, Perplexity-User
• Google-Extended (Gemini training)
• Applebot-Extended (Apple Intelligence)
• Bingbot (Bing Copilot)
• CCBot (Common Crawl — feeds many AI projects)
• cohere-ai
• Meta-ExternalAgent

Blocking any of these means opting out of the AI engine that uses it. A few brands have content-licensing reasons to block; almost no one else should.

Step 9 — Build brand co-citation across the open web

The trickiest of the nine — but the most defensible. AI engines learn brand–topic associations from co-occurrence across many domains. If 50 unrelated sites mention “Frostbite Marketing” in articles about “digital marketing agency”, the engines learn the association.

Sources of co-citation:

• Press releases published on major newswires
• HARO / Featured.com / Qwoted responses that land in published articles
• Podcast guest appearances
• Awards and rankings (Inc 5000, Clutch, G2, industry-specific lists)
• Expert quotes in roundup posts
• Industry directories with editorial content
• Conference speaking and the post-event coverage

Each adds one more co-citation. Over 12 months, brands with active citation pipelines build a moat that’s hard for competitors to catch.

Timeline: when to expect results

Realistic expectations for a brand starting fresh:

• Week 1–4: Technical foundation in place. Schema validated. llms.txt live. AI crawlers unblocked.
• Month 2–3: First Perplexity and Bing Copilot citations on long-tail queries.
• Month 3–6: First Google AI Overview appearances (lags classic ranking by 1–2 months typically).
• Month 6–9: ChatGPT and Claude start surfacing the brand in unprompted answers (training cutoff dependent — varies by model).
• Month 9–18: Brand co-citation compounds; AI citation rate accelerates non-linearly.

There’s no shortcut to month 9. There’s also no upside to delaying the start.

Common mistakes that prevent AI citation

• Blocking AI crawlers — see Step 8
• Stale dates — see Step 5
• Anonymous content — see Step 6
• No llms.txt — see Step 7
• Marketing-fluff opening paragraphs — see Step 2
• Inconsistent facts across pages — AI engines penalize contradictions
• Identical content across location pages — devalues the whole site, not just the duplicates
• Schema mistakes — incorrect or missing schema can cause AI engines to skip the page entirely
• Slow Core Web Vitals — performance issues reduce both classic ranking AND AI eligibility

How Frostbite Marketing builds AI citation pipelines

For our 250+ clients across 10+ states, we treat AI citation as a managed program:

• Foundation setup in weeks 1–4 (schema, llms.txt, crawler config, author profiles, freshness signals)
• Content production that bakes in answer-first formatting, question H2s, and FAQ schema automatically — see our Content Marketing service
• Citation pipeline (press releases, Featured.com, podcast outreach, awards submissions) to drive co-citation breadth
• Monthly AI engine monitoring — we prompt ChatGPT, Claude, Perplexity, AI Overviews, and Bing Copilot on your branded and category queries, and track citation rate over time
• Quarterly content refresh on cornerstone pages to maintain freshness signals

The compound effect: brands we manage typically see AI citation rates double in 6 months and triple by month 12.

FAQ