The GEO Content Playbook: Data-Driven Strategies That Earn AI Citations in 2026

We've been tracking AI citation patterns across ChatGPT, Perplexity, and Google AI Overviews for the past year — analyzing which of our clients' pages get cited and which get ignored. The pattern is consistent: content that follows specific, measurable formatting conventions earns citations at dramatically higher rates than content that doesn't, regardless of domain authority.

There are now over 1.5 billion weekly active users across ChatGPT, Gemini, and Perplexity combined. These platforms don't rank websites — they cite sources. And the Princeton/Georgia Tech GEO research paper (arXiv:2311.09735) proved what we'd been seeing in practice: quotation addition improved visibility by 42.8%, statistics by 33.2%, and source citations by 27.8%. Meanwhile, keyword stuffing — the old SEO staple — decreased AI visibility by 8.7%.

This playbook translates that research into a practitioner's framework you can implement this week. If you're new to Generative Engine Optimization, start with our introduction to GEO first. If you need the technical infrastructure layer (robots.txt, llms.txt, schema), see our AEO technical stack guide.

The GEO Research: What Actually Earns AI Citations

Generative Engine Optimization (GEO) is the practice of structuring content so AI platforms cite it in generated answers. The Princeton GEO study (arXiv:2311.09735) tested nine optimization methods and found that statistics (+33.2%), quotations (+42.8%), and source citations (+27.8%) produce the largest measurable improvements in AI visibility.

The 9 Methods, Ranked by Impact

Here are the nine content optimization methods tested in the GEO research, ordered from highest to lowest impact on AI visibility:

Three patterns dominate: quotations, statistics, and source citations. If you only optimize for three things, optimize for these. Combined, they represent over 100 percentage points of potential visibility improvement.

Why Quotations, Statistics, and Citations Work

AI language models are trained to identify and prioritize well-sourced information. When a model encounters a claim backed by a statistic from a named source, it can verify that claim against its training data. When it finds an expert quote, it has an attributable piece of information it can reference without risk of hallucination. And when content explicitly cites external sources, the model treats it as a node in a larger knowledge graph — not an isolated opinion.

As Rand Fishkin, co-founder of SparkToro, observed: "AI systems don't just want answers — they want answers they can trace back to authoritative sources. The age of unsourced content marketing is over." This mirrors what we see in practice: pages with 3+ inline citations per 1,000 words receive significantly more AI referral traffic than pages with zero citations.

The Answer Capsule: The Most Important Content Pattern for AI

An answer capsule is a 40-60 word direct answer placed immediately after an H2 heading. It is the single most extractable content pattern for AI systems. When an AI engine encounters a question-format heading followed by a concise, self-contained answer, it can lift that paragraph nearly verbatim into its response.

How to Write an Answer Capsule

Every answer capsule follows the same three-sentence structure:

1. Definitional sentence— Start with a clear, factual statement that includes the target concept in the first 10 words. Example: "Generative engine optimization (GEO) is the practice of structuring content so AI systems cite it."
2. Context sentence— Add one sentence of elaboration that distinguishes the concept, provides a key comparison, or adds a relevant data point. Example: "Unlike traditional SEO which targets ranking positions, GEO focuses on citation frequency and prominence in AI-generated responses."
3. Impact sentence— Close with a specific number, outcome, or implication. Example: "Research shows GEO optimizations can boost AI visibility by up to 40%."

Before and After Examples

Before (generic blog intro):

In today's fast-paced digital landscape, businesses are increasingly looking for ways to improve their online presence. One area that has gained a lot of attention recently is how AI search engines work and how companies can appear in AI-generated results. Let's dive into what this means for your business.

After (answer capsule):

Generative Engine Optimization (GEO) is the practice of structuring website content so AI platforms like ChatGPT, Perplexity, and Google AI Overviews cite it as a source in their generated responses. Unlike traditional SEO which optimizes for search result rankings, GEO targets citation frequency — how often and how prominently your content appears in AI answers. According to the Princeton GEO study, sites implementing GEO best practices see up to 40% improvement in AI visibility metrics.

The second version is self-contained, starts with a definition, includes a comparison for context, and closes with a specific data point. An AI system can extract it as a complete, attributable answer.

How to Format Statistics That AI Systems Actually Cite

Statistics addition improved AI visibility by 33.2% in the GEO study — making it the second-highest-impact optimization method. But not all statistics are created equal. AI systems preferentially cite statistics that include the source organization, publication year, sample size, and a specific numerical value. Vague claims like "most businesses see improvement" are ignored.

Inline Statistics (Best for AI Extraction)

The highest-cited format embeds statistics directly within flowing prose, with source attribution in the same sentence:

According to a 2026 Gartner forecast, 25% of traditional search engine traffic will shift to AI chatbots and virtual agents by end of 2026, with that figure projected to reach 50% by 2028. Early adopters of GEO strategies report that AI referral leads convert at 25x the rate of traditional organic search leads (Go Fish Digital, 2025).

Statistics Blocks (For Scanning)

When you have multiple related data points, group them into a scannable block with bold key metrics and a source line:

• Hundreds of millions of weekly active users on ChatGPT, with OpenAI reporting rapid growth throughout 2025-2026
• Billions of users exposed to AI answers via Google Gemini and AI Overviews across Google Search
• Tens of millions of monthly users on Perplexity, making it the fastest-growing AI search platform
• 58.5% of Google searches in the US end without a click to any website (SparkToro/Datos, 2025 zero-click study)
• Significant CTR declineswhen Google AI Overviews appear — Semrush's 2025 study found organic click-through rates drop substantially for queries with AI Overviews

Rules for Citable Statistics

1. Always include the source name and date— "67% of businesses use AI tools (Gartner, 2026)" not "67% of businesses use AI tools"
2. Include sample size when available— "a survey of 1,200 enterprises" adds credibility
3. Use specific numbers— "42% reduction" not "significant reduction"
4. Bold the key metric — visual hierarchy helps both humans and AI parsers
5. Group related statistics together — AI systems extract data clusters more effectively than scattered individual stats
6. Prefer recent sources — AI systems strongly favor recent data, with Perplexity and ChatGPT both showing clear recency bias in their citation patterns

Platform-Specific GEO Strategies: ChatGPT vs. Perplexity vs. Gemini vs. AI Overviews

Each AI platform weights different signals when selecting sources to cite. ChatGPT favors domain authority and freshness via Bing's index, Perplexity prioritizes recency and direct sourcing, Gemini leans on Google's Knowledge Graph, and AI Overviews cite from top-ranking Google results. Effective GEO means optimizing for each platform's specific citation logic.

ChatGPT: Domain Authority + Freshness

ChatGPT uses a retrieval-augmented generation (RAG) system powered by Bing's search index when browsing is enabled. Analysis by SE Ranking suggests that the vast majority of ChatGPT citations come from recently updated content — typically within the last 10-12 months. Pages with clear publication timestamps also appear to receive significantly more citations than undated pages.

What ChatGPT favors:

• Domain authority — high-DR sites still get preferential citation. ChatGPT draws heavily from established publications and well-linked websites.
• Content freshness — visible datePublished and dateModified timestamps in both the markup and the page content.
• FAQ sections — pages with structured FAQ sections nearly double their chances of being cited by ChatGPT for question-based queries.
• Comprehensive coverage — ChatGPT prefers citing a single authoritative page over stitching together information from multiple thin pages.

Perplexity: Recency + Specificity

Perplexity is the most citation-visible AI platform — it shows numbered source links directly in its responses. With over 45 million monthly active users, it's where your GEO efforts will show the most measurable results.

What Perplexity favors:

• Extreme recency— content published or updated within the last 3-6 months gets cited 3-4x more frequently than older content. Perplexity's index refreshes faster than ChatGPT's.
• Specific, factual content— Perplexity strongly prefers pages with concrete data points over opinion pieces. A page stating "the average cost is $5,000-$50,000" beats "costs vary depending on several factors."
• Reddit and forum presence — Perplexity indexes Reddit heavily. Answering questions in your niche subreddits with links to your detailed content creates citation pathways.
• Structured comparisons — tables comparing tools, services, or approaches are extracted at a significantly higher rate than flowing prose.

Google Gemini + AI Overviews: E-E-A-T + Schema

Google AI Overviews now appear for approximately 13% of US search queries (Semrush, October 2025), expanding beyond purely informational queries into commercial and transactional searches. Because AI Overviews pull from Google's search index, traditional SEO signals remain highly relevant.

What Google AI Overviews favor:

• E-E-A-T signals — author credentials, editorial policies, organizational authority. Google has the most sophisticated quality assessment system of any AI platform.
• Schema markup— Article, FAQPage, HowTo, and BreadcrumbList schemas help Google's AI understand your content structure. For implementation details, see our AEO technical stack guide.
• List and step formats — numbered steps and bulleted lists are extracted at a much higher rate than paragraphs for how-to queries.
• Topical authority — sites with deep coverage of a topic (hub + spoke content clusters) get cited more than sites with a single page on a topic. Our topical authority guide covers this in depth.

Platform Signal Comparison

6 Content Patterns That Earn the Most AI Citations

Beyond answer capsules and statistics, six content structures earn disproportionately more AI citations: definitions, step-by-step processes, comparison tables, data-backed lists, framework introductions, and FAQ sections. These patterns are structurally extractable — AI systems can parse them into direct answers without reformulating surrounding prose.

1. Definitions and Glossary Entries

Definitions are the most-extracted content type by AI. Every definition query ("what is X") requires the model to find and cite a clear, authoritative source. The winning format:

[Term] is [concise 1-sentence definition]. [1-2 sentences of elaboration with context, origin, or key distinction]. [1 sentence with a specific example or application].

If you haven't already, create a comprehensive glossary for your industry. Our SEO glossary covers 100+ terms and is one of our most-cited resources by AI platforms.

2. Structured Comparisons

"X vs Y" and "which is better" queries are among the most common in AI search. AI platforms cite comparison tables at a higher rate than prose comparisons because tables provide clean, extractable data.

Table rules for AI extraction:

• Use clear header rows with descriptive column names
• Keep cell content concise (under 50 characters per cell)
• Use consistent units within columns
• Include a descriptive heading above the table
• Close with a clear recommendation paragraph starting with "Our recommendation:"

3. Step-by-Step Guides

How-to content with numbered steps gets extracted at significantly higher rates than unstructured tutorials. The key is making each step self-contained and starting with an action verb:

• Step 1: Audit existing markup— Use Google's Rich Results Test to identify schema gaps.
• Step 2: Add Organization schema — Insert JSON-LD on your homepage with sameAs links.
• Step 3: Implement Article schema — Every content page needs Article markup with author details.

4. Expert Quotations

Quotation addition was the single highest-impact GEO method at +42.8% visibility. AI systems treat attributed quotes as high-confidence information because they come from named, verifiable sources.

Effective quote formats:

• "According to [Name], [Title] at [Organization]..." — full attribution chain
• "As [Name] noted in [Publication/Event/Interview]..." — adds a verifiable source
• "[Name], who has [credential/experience], explains: '...'" — establishes authority before the quote

The quote doesn't have to be from a celebrity. Industry practitioners, researchers, and even your own team members with documented expertise all work — as long as the attribution is specific and the person is verifiable online.

5. Lists and Rankings

"Best X for Y" content is heavily queried across all AI platforms. Structure these with a numbered list, a clear evaluation framework, and specific pros/cons for each item. Always state your evaluation criteria upfront so the AI (and reader) understands why items are ranked the way they are.

6. FAQ Compilations

FAQ sections serve double duty: they provide extractable Q&A pairs for AI systems, and they support FAQPage schema markup that makes the content machine-readable. AI search queries average 23 words (compared to 4 words on traditional Google search) — they're essentially questions. Your FAQ section should mirror these natural language questions.

FAQ rules for AI extraction:

• Use the full natural-language question as the H3 heading
• Start the answer with a direct response — never "Great question!" or "It depends"
• Include specific numbers, ranges, or timeframes in every answer
• Keep each answer to 50-150 words
• Add FAQPage schema markup to the section

The Citation-Optimized Writing Style

Citation-optimized writing is a content style designed to maximize the probability that AI systems extract and attribute your content. It combines authoritative tone, specific data, and clear structure to create text that language models identify as high-confidence, citable information. Based on the GEO research findings, this style increases AI visibility by 12-43% depending on the specific technique applied.

7 Rules for Citation-Optimized Content

1. Lead with authority, not hedging— Write "According to our analysis of 500 implementations..." not "We think...". Definitive statements with evidence are cited; hedged opinions are ignored.
2. Use specific numbers everywhere— "Reduces response time by 42%" not "significantly reduces response time." Every claim should have a number attached to it.
3. Attribute quotes with full context— "As Dr. Andrew Ng of Stanford noted at NeurIPS 2025..." The more verifiable the attribution, the more likely the citation.
4. Cite external sources inline— "(Source: Gartner, 2026)" or "(McKinsey Global Survey, n=1,200)" after every data point. Never present statistics without attribution.
5. Use technical terms correctly— Domain-specific vocabulary signals expertise. Write "retrieval-augmented generation (RAG)" not "AI that looks things up." The GEO study showed technical terms improve visibility by 18.5%.
6. Be definitive in recommendations— "The best approach is X because Y" not "X might be a good approach." AI systems cite decisive content over equivocating content.
7. Front-load the answer, then elaborate — Every section should start with the conclusion, then provide supporting evidence. This inverted pyramid structure ensures the most important information gets extracted even if the AI only reads the first paragraph.

The GEO Audit: How to Check Your Current AI Visibility

A GEO audit measures how your brand currently appears (or doesn't) across AI platforms. Before optimizing anything, you need a baseline. This is a systematic process of querying each major AI platform with your target topics and documenting where you appear, how you're described, and what competitors get cited instead.

Manual Audit Protocol (Weekly)

Run these queries across ChatGPT, Perplexity, Google Gemini, and Claude for each of your core topics:

1. Brand queries— "What is [Your Company]?", "Tell me about [Your Product]"
2. Category queries— "What are the best [your service] companies?", "Top [your product category] in 2026"
3. Problem queries— "How do I [solve the problem your product addresses]?"
4. Comparison queries— "[Your Product] vs [Competitor]", "[Competitor] alternatives"

For each query, record: Were you mentioned by name? Were you cited as a source with a link? What was the sentiment? Which competitors appeared? What specific information was attributed to you?

Automated Tracking Tools

Proxy Metrics When You Don't Have Dedicated Tools

Even without specialized GEO tracking tools, monitor these proxy metrics that correlate with AI citation performance:

• Branded search volume (Google Search Console) — rising brand searches often correlate with AI mentions driving people to verify claims.
• Direct traffic (GA4) — AI citations drive users directly to your site. An unexplained rise in direct traffic often indicates AI referrals.
• Referral traffic from AI platforms — check for chat.openai.com, perplexity.ai, and gemini.google.com in your referral sources.
• "According to [brand]" searches — people verify claims they read in AI responses. Track this phrase in Search Console.
• Backlink velocity — AI-cited content often attracts more backlinks as people reference the same sources the AI did.

Measuring GEO Results: The NORA Framework

The NORA framework evaluates content across four dimensions that predict AI citation probability: Notability, Objectivity, Recency, and Authority. Every piece of content should score high on all four to maximize AI visibility.

The Four Dimensions

• Notability— Does this cover something people actually search for? Validate with keyword research and AI query analysis. Content about topics nobody asks AI about won't get cited regardless of quality.
• Objectivity — Is the content balanced and fact-based? AI systems deprioritize overtly promotional or biased sources. Present both sides of comparisons. Acknowledge limitations of your own products. Let data speak.
• Recency — Is the content recently published or meaningfully updated? Visible timestamps matter. Update statistics every 3-6 months. Add new sections when the topic evolves. Remove outdated references.
• Authority — Does the author have verifiable expertise? Link to author profiles with credentials, LinkedIn, and published work. Our E-E-A-T implementation guide covers this in full.

GEO Content Scorecard

Before publishing any content, run it through this checklist:

• Does every H2 section start with a 40-60 word answer capsule?
• Are there 5+ sourced statistics per 2,000 words?
• Does the content include at least 2 expert quotations with full attribution?
• Are there structured tables or comparison elements?
• Is there a FAQ section with FAQPage schema?
• Does the content use domain-specific technical terms correctly?
• Is the author identified with credentials and a linked profile?
• Is datePublished and dateModified visible on the page?
• Are all claims sourced with (Organization, Year) attribution?
• Is the content 2,000+ words with comprehensive topic coverage?

Real-World GEO Results: What the Data Shows

The GEO research paper provides the theory. But what does this look like in practice? Go Fish Digital reported that their GEO-optimized content strategy generated significantly more qualified leads from AI search channels within 6 months. The key changes they documented: adding answer capsules to every H2 section, including 5+ sourced statistics per page, and implementing FAQPage schema across all content pages.

We've seen similar results with our own content. Pages on our SEO blog that follow the GEO content patterns outlined in this playbook receive measurably more AI referral traffic from perplexity.ai and chat.openai.comthan pages that don't. The pattern is clear: pages with structured FAQ sections, sourced statistics, and answer capsules outperform pages with only long-form prose — even when the prose pages are more comprehensive overall.

Additional data points from industry research:

• AI referral leads convert at substantially higher rates than traditional organic search leads, according to Go Fish Digital's analysis — likely because users arriving from AI citations have higher intent and pre-established trust.
• Early GEO adopters report that a growing share of sales-qualified leads now originate from AI search channels, with some companies reporting 20-30%+ (SEO.com, 2026).
• Faster pages correlate with more AI citations — research from Lattice Ocean (2025) found that pages with sub-second FCP received significantly more AI citations than slower pages. Use our free page speed estimator to check your current performance.
• Broad, topic-aligned titles get 2x more citations than keyword-stuffed titles (SE Ranking, 2025). Use our headline analyzer to evaluate your titles.

Entity Building: Getting AI to Recognize Your Brand

AI systems cite entities they can verify across multiple independent sources. If your brand only exists on your own website, language models have low confidence in citing you. Entity building is the practice of establishing your brand across platforms that AI systems use for cross-referencing — creating a web of verifiable information that models can triangulate.

Entity Verification Priority List

1. Wikipedia — The number-one entity verification source for AI. If your brand is notable enough, a Wikipedia article is the single most impactful entity signal. Even a Wikidata entry (without a full Wikipedia article) helps establish your entity in AI knowledge graphs.
2. Crunchbase — Claim your company profile with accurate founding date, team, funding, and description.
3. Google Business Profile — Claim and optimize even for online-only businesses.
4. LinkedIn Company Page — Complete all fields and post regularly.
5. Industry directories — G2, Product Hunt, Clutch, and vertical-specific directories.
6. Schema.org sameAs — Link all these profiles from your Organization schema so AI systems can verify the connections.

Multi-Platform Presence

AI systems aggregate information from across the web. Your presence beyond your website directly impacts citation probability:

• Reddit — AI platforms (especially Perplexity) index Reddit heavily. Answer questions in your niche subreddits with genuine expertise and link to your detailed content.
• YouTube — Video transcripts get indexed by AI. Create content that restates your key expertise in video format.
• GitHub — For technical brands, open-source contributions build authority that AI systems recognize.
• Guest posts on industry publications — Bylined articles on authoritative sites create additional citation pathways. The AI sees your name and expertise on a trusted domain.
• Podcast appearances — Transcripts get indexed. Mention your key data points and frameworks by name.

Content Freshness: The Overlooked GEO Factor

Content freshness is the most underrated GEO signal. Perplexity weights recency so heavily that content published within the last 3-6 months gets cited 3-4x more than older content, even when the older content is more comprehensive. ChatGPT shows a similar pattern: 95% of its citations come from content updated within 10 months.

The Freshness Playbook

1. Display "Last updated: [date]" visibly on every page — not just in the schema, but in the actual visible content. AI systems read visible dates.
2. Refresh statistics every 3-6 months — replace 2024 data with 2025 data, replace 2025 data with 2026 data. Stale statistics are the fastest way to lose AI citations.
3. Add new sections when topics evolve— don't just update old sections. Add entirely new H2 sections that cover recent developments.
4. Update dateModified in your schema— but only when you've made a substantive content change. Search engines and AI systems detect fake timestamp updates.
5. Re-submit updated URLs via Google Search Console — trigger recrawling after meaningful updates.

Your GEO Content Implementation Checklist

A minimum-viable GEO implementation covers four phases over 30 days: answer capsules and statistics on your top 20 pages (week 1), FAQ sections and schema markup (week 2), platform-specific optimizations (week 3), and measurement setup with ongoing content freshness (week 4). Complete these items on high-traffic pages first, then expand site-wide.

Do This Week (High Impact, Low Effort)

1. Add a 40-60 word answer capsule after every H2 heading on your top 10 pages
2. Add 3+ sourced statistics to each page (with source name, year, and sample size)
3. Add 1-2 expert quotations per page with full attribution
4. Make datePublished and dateModified visible on every content page
5. Update the dateModified value and schema when making these changes

Do Next Month (Medium Effort)

1. Run a full GEO audit across ChatGPT, Perplexity, Gemini, and Claude for your top 20 queries
2. Create FAQ sections with FAQPage schema on all service and product pages
3. Build comparison tables for every "vs" and "best" query in your niche
4. Set up weekly AI platform monitoring (manual or with tools)
5. Implement the technical AEO stack: robots.txt AI crawler rules, llms.txt, and schema markup — see our complete AEO technical guide

Do Ongoing (Monthly)

1. Update statistics and data points on all major content pages
2. Publish 1+ piece of original data, research, or survey results
3. Monitor AI referral traffic in GA4 (perplexity.ai, chat.openai.com)
4. Test brand queries on AI platforms and document changes
5. Review and update content that hasn't been modified in 6+ months

Frequently Asked Questions

What is Generative Engine Optimization (GEO)?

Generative Engine Optimization (GEO) is the practice of structuring website content so AI platforms like ChatGPT, Perplexity, and Google AI Overviews cite it when generating answers. Unlike traditional SEO which optimizes for search result rankings, GEO optimizes for citation frequency and prominence within AI-generated responses. The Princeton GEO study found that specific content patterns — including statistics, expert quotations, and source citations — can improve AI visibility by 27-43%.

Does GEO replace traditional SEO?

No. GEO and traditional SEO are complementary strategies that reinforce each other. Traditional SEO signals like domain authority, backlinks, and E-E-A-T directly influence AI citation probability — especially for ChatGPT (which uses Bing's index) and Google AI Overviews (which use Google's index). The best approach is to layer GEO content optimizations on top of a strong SEO foundation. For a detailed comparison, see our GEO vs. traditional SEO analysis.

How long does it take to see GEO results?

Initial results typically appear within 2-4 weeks for Perplexity (which re-indexes frequently) and 4-8 weeks for ChatGPT and Google AI Overviews. However, building sustainable AI citation authority takes 3-6 months of consistent content optimization. The fastest way to see results is to update your top-performing existing pages with answer capsules, statistics, and quotations — new pages take longer to build authority from scratch.

How much does GEO cost?

GEO is primarily a content optimization strategy, so the main cost is time rather than tools. Most GEO optimizations — answer capsules, statistics addition, FAQ sections — can be applied to existing content with no additional tools. Dedicated AI visibility tracking tools like Semrush AI Visibility (included in Semrush plans) or Scrunch AI (free tier available) add monitoring capability. For a free start, use our Neural Audit tool to analyze your current AI search visibility at no cost.

Does keyword density matter for GEO?

Keyword stuffing actively hurts GEO performance. The Princeton study found it decreases AI visibility by 8.7% — the only method with a negative impact. AI systems evaluate semantic relevance and topical coverage, not keyword frequency. Write naturally using domain-specific technical terms, and let keyword usage emerge from comprehensive topic coverage rather than forced repetition. Our keyword density checker helps verify you're not over-optimizing.

Can small businesses compete in GEO?

Yes — and in some ways, small businesses have an advantage. AI platforms cite niche expertise heavily. A small business that publishes deep, data-rich content in a specific niche can outperform large generalist sites for topic-specific queries. The key is topical authority: cover your niche comprehensively rather than trying to compete on broad topics. Start with a glossary, FAQ pages, and how-to guides for your specific industry.

The Bottom Line

GEO is not a separate discipline from SEO — it's an evolution of it. The same fundamentals apply: create comprehensive, authoritative content that serves user intent. The difference is in execution. AI systems need structured, extractable, well-sourced content they can cite with confidence. Answer capsules, statistics with attribution, expert quotations, and FAQ sections aren't just nice-to-have formatting — they're the signals that determine whether your content gets cited or ignored.

Start with your top 10 pages. Add answer capsules, source your statistics, include expert quotes, and build FAQ sections. Then implement the AEO technical stack to ensure AI crawlers can find and understand your content. The businesses that build this foundation now will own the citations in the AI search era.

Not sure where to start? Run a free SEO audit on your site to identify technical gaps, then use the Neural Audit to establish your AI visibility baseline. From there, apply the answer capsule format to your highest-traffic pages first and work outward. Most sites see initial Perplexity citation improvements within 2-4 weeks of implementing these changes.