AI SEO Strategy: A Step-by-Step Playbook for 2026

ChatGPT, Perplexity, and Google’s AI Overviews now answer questions your buyers used to type into a search box. If your site isn’t structured for machines to read, quote, and trust, you don’t lose a ranking. You disappear from the answer entirely. An AI SEO strategy is how you get back in the room: a sequence of technical, content, and authority fixes that make your site legible to AI crawlers and citable by AI answers.

This playbook lays out that sequence across 90 days: audit, quick wins, content optimization, authority building, measurement, and a maintenance loop you repeat every quarter. Run it in order. Each phase depends on the one before it.

The 90-Day Roadmap at a Glance

What an AI SEO Strategy Actually Covers

An AI SEO strategy is a plan for making your content discoverable, parseable, and citable by AI systems like ChatGPT, Claude, Perplexity, and Google’s AI Overviews, not just by traditional search rankings. It shares roots with SEO for AI search engines and generative engine optimization, but it adds a layer classic SEO never had to handle: bot-level access control and machine-readable structure.

Traditional SEO asks whether Google will rank your page. AI search optimization asks an earlier question: can an AI crawler even reach the page, and once it does, can the model extract a clean, citable answer from it? Backlinks and keyword density still matter, but they’re no longer sufficient on their own. A page can rank on page one and still be invisible in an AI answer if a crawler is blocked, the content loads behind JavaScript, or the actual answer is buried under three paragraphs of throat-clearing.

That’s the gap this playbook closes, phase by phase: technical access first, then AI content optimization, then the authority signals that turn a citable page into a repeatedly cited one.

Week 1–2 Audit: Find Your AI Visibility Baseline

The audit phase measures where your site currently stands with AI crawlers and AI-generated answers before you change anything. Skip it and you’ll spend the next 11 weeks fixing problems you don’t actually have while the real ones sit untouched.

Start with a structured baseline instead of guesswork. ICODA’s free AI Visibility Checker scores a domain across four areas: how many people ask AI about your market each month, whether you actually show up in live AI answers against competitors, a ranked list of channels to fix first, and the technical layer underneath it all: bot access, schema coverage, and llms.txt status.

Audit checklist:

• Run your domain through an AI visibility checker and record the baseline score
• Pull server logs and confirm GPTBot, ClaudeBot, PerplexityBot, and Google-Extended can reach your key pages
• List every schema type currently live on the site (or confirm there’s none)
• Query ChatGPT, Claude, and Perplexity directly with 10–15 buyer questions and note whether your brand appears, and how it’s described
• Check whether your most important pages render without a login, a paywall, or heavy client-side JavaScript

Write the baseline down. You’ll compare against it in the measurement phase, and “we improved” only means something next to a number.

Week 2–4 Quick Wins: robots.txt, llms.txt, Basic Schema

Quick wins fix the access layer: the technical settings that decide whether AI crawlers can reach your content at all before content quality even enters the picture. Most of this ships in days, not weeks, and it’s the highest-leverage work in the entire playbook because nothing downstream matters if the crawler never gets in the door.

Fix robots.txt first. The instinct to block every AI bot is usually wrong. Training crawlers and search-time crawlers are separate, and blocking one doesn’t block the other. Rutgers and Wharton researchers found that publishers who blocked AI crawlers outright saw traffic drop by more than 23% without any reliable drop in citations elsewhere, because the block hit search-time bots too. A selective posture works better for most sites:

Verify each directive against server logs, not just the file itself. Some bots, Perplexity’s undeclared crawlers among them, have been documented ignoring robots.txt entirely, so treat the file as a strong signal to well-behaved bots and back it with server-level rules for anything that doesn’t comply.

Ship an llms.txt file, but keep expectations honest. This is a plain-text file at your domain root that points AI systems to your most important pages. Independent studies in 2026 put adoption at roughly 9–10% of major domains, and most AI search crawlers, GPTBot, ClaudeBot, and PerplexityBot included, mostly skip fetching it when they’re building a citation. Where it does earn its keep today is with coding agents: tools like Cursor, Claude Code, and GitHub Copilot fetch llms.txt routinely on documentation sites. For a content-driven brand, treat it as a half-day, low-cost bet on a standard that may matter more later, not a guaranteed traffic lever.

Add basic schema. At minimum, implement Organization schema (name, logo, sameAs links to your verified social and review profiles) and Article schema on every blog post (author, publish date, headline). This is the foundation everything in the content optimization phase builds on.

Quick wins checklist:

• Rewrite robots.txt with a selective posture for training vs. search-time bots
• Confirm the new rules in server logs within a week
• Publish llms.txt with links to your 10–20 most important pages
• Add Organization and Article schema sitewide
• Re-run the visibility checker to confirm bot access improved

Month 2 Content Optimization: Headings, FAQ Markup, Entity Building

Content optimization restructures what’s already on the page so an AI model can extract a complete answer without stitching together fragments from three different paragraphs. This is where AI content optimization stops being a technical exercise and becomes an editorial one.

Fix your heading hierarchy first. Write each H2 as the question a reader (or a model) would actually ask, then answer it in the first one to two sentences underneath. Save context, caveats, and nuance for after the answer, not before it. This single change does more for AI Overview eligibility than almost any other content edit, because it matches exactly how these systems extract passages: a self-contained chunk that answers a question completely on its own.

Be honest about FAQ markup. Google deprecated FAQ rich results from Search on May 7, 2026, so the expandable Q&A dropdown is gone from the SERP. FAQPage schema itself is still a valid Schema.org type, and Google has said unused structured data doesn’t hurt a page. What it hasn’t said is that FAQ schema improves AI citation odds. Treat it as one more layer of machine-readable structure on genuinely useful Q&A content, not as a growth hack. If a FAQ section only exists to house schema, cut it. If it answers real questions users ask, keep the markup and don’t expect it to do more than it can.

Build entities, not just pages. AI models cite sources they can identify with confidence. That means your brand name, your authors, and your core topics need to appear consistently across your site, your social profiles, and third-party mentions, not just inside your own content. Practical steps:

• Give every article a named author with a bio page and consistent credentials
• Link author and organization profiles with sameAs schema to LinkedIn, X, and industry directories
• Use the same entity names and descriptions across your site, your Google Business Profile, and any Wikidata or Wikipedia presence you qualify for
• Interlink related articles into topic clusters instead of leaving them isolated

Content optimization checklist:

• Rewrite H2s as questions with a direct answer in the first sentence
• Audit existing FAQ schema and remove it from pages with no real FAQ content
• Publish an author bio page for every contributor
• Add sameAs links across all author and organization schema
• Build or tighten topic clusters through internal linking

Want To Dive Deeper?

ICODA offers a full SEO/AI SEO audit for free

Get Your Audit

Month 3 Authority Building: Structured Data, Citations, Expert Mentions

Authority building earns the third-party signals AI models use to decide whether your brand is a trustworthy source, on top of the technical structure you already fixed. Backlinks still count here, but AI citation research increasingly points to a wider set of signals: Wikipedia and Wikidata presence, mentions on Reddit and industry forums, and reviews on independent platforms.

Expand structured data beyond the basics. Add BreadcrumbList schema for navigation clarity, Review or Product schema where relevant to your business, and keep every schema type in sync with what’s actually visible on the page. Mismatched schema, markup that claims something the page doesn’t show, is a common source of AI models getting your brand wrong.

Earn citations deliberately. Digital PR, original research, and data-backed guest content still work, but the target has widened. Getting quoted in a niche industry newsletter or cited correctly on a comparison thread on Reddit can matter as much to AI visibility as a traditional backlink, because these are exactly the sources AI models pull into their training and retrieval systems.

Put a name and a credential on the work. E-E-A-T requirements now extend past finance and health content to nearly every category. A named author with real domain experience, an original data point, or a direct quote from a practitioner gives an AI model something concrete to attribute, instead of a vague “industry experts say.”

This is also where a dedicated AI SEO service earns its cost for teams without the in-house bandwidth to run citation outreach alongside the technical work.

Authority building checklist:

• Add BreadcrumbList and any relevant Review/Product schema
• Audit all schema for mismatches against visible page content
• Pitch 3–5 pieces of original data or commentary to industry publications this quarter
• Claim or correct your brand’s Wikidata entry if one exists
• Add named author credentials to every piece of cornerstone content

Measurement Framework: What to Track and How

Measuring AI SEO tracks whether AI systems actually cite you, not just whether your rankings moved, because the two increasingly diverge. A page can hold position four in Google and never appear in a single AI Overview, or the reverse.

Referral traffic from AI platforms is still small on most sites, so don’t treat a flat number as failure on its own. Citation frequency and share of voice are the leading indicators; referral traffic is the lagging one that catches up once citations compound.

Quarterly Maintenance Checklist

AI SEO maintenance re-runs the audit on a fixed schedule, because bot behavior, schema requirements, and AI Overview mechanics all shift faster than a typical SEO refresh cycle. A configuration that worked in Q1 can quietly break by Q3 if a new crawler launches and your robots.txt never gets updated.

• Re-run the AI Visibility Checker and compare against your baseline
• Check for new AI crawler user-agents and add explicit rules for them
• Refresh any content older than 12 months on a topic where freshness matters
• Re-validate all schema against current markup
• Re-query your core buyer questions across ChatGPT, Claude, and Perplexity
• Review which competitors gained or lost citation share since the last check

Building an AI SEO Strategy That Compounds

None of these phases work in isolation. Fixing robots.txt without restructuring content gets crawlers in the door with nothing worth citing behind it. Rewriting content without fixing bot access means a well-structured answer that a crawler never reaches. The 90-day sequence exists because access, structure, and authority build on each other in that order, not because the calendar demands it.

Run the audit again in 90 days and you’ll have something most competitors still don’t: an actual number to point to instead of a hunch about whether AI search optimization is working. That number is the real deliverable of an AI SEO strategy, and it’s the one that keeps the next quarter’s roadmap grounded in evidence instead of guesswork.

Most sites still treat SEO for AI search engines as an afterthought bolted onto their existing content calendar. The ones that treat it as its own 90-day sequence, with a baseline, a build phase, and a repeat cycle, are the ones that show up when a buyer asks an AI assistant instead of typing into Google.