In November 2024, a B2B SaaS client noticed something alarming. Their top blog posts — pieces that had driven 4,000-8,000 monthly visits for years — had dropped 30-45% in traffic. Rankings hadn’t changed. The pages were still in positions 1-3 for their target keywords.
The visits simply weren’t coming.
The culprit was Google AI Overviews. Their content was being summarised and answered directly on the results page. Users were getting the information without ever clicking.
This is the environment AEO exists to address. Not ranking higher. Being selected as a cited source.
What AEO Actually Is
Answer Engine Optimization is the practice of structuring content so that AI systems — Google AI Overviews, ChatGPT, Perplexity, Claude — extract and cite it when answering user queries.
This is different from traditional SEO in one critical way. Traditional SEO is about ranking position. A page in position 2 gets a predictable percentage of clicks relative to
position 1. The game is about moving positions.
AEO is binary. When an AI Overview appears for a query, there are typically 3-8 cited sources. Those sources receive substantial traffic. Every other page that ranks for that query receives essentially nothing from that query instance. You’re either cited or you’re irrelevant.
Semrush’s December 2025 analysis found AI Overviews appearing on 47% of informational English queries. For a brand whose traffic is primarily informational (how-to guides, explainers, educational content), this binary dynamic applies to nearly half their total search volume. The question is no longer “what position do I rank?” It’s “do I get cited at all?”
How AI Overview Citation Works: The Mechanics
Understanding what determines citation requires understanding how Google’s AI Overviews select sources.
The system is not simply pulling the top-ranked pages for a query. It’s running a retrieval-augmented generation process: it identifies candidate pages, evaluates them for relevance and quality, extracts passages that directly answer the query, and synthesises those passages into a coherent response. The pages it extracts from become the cited sources.
The selection criteria involve multiple factors we’ve been able to identify through systematic testing:
Directness of answer. AI systems strongly prefer content that answers the specific question in the first 2-3 sentences of a section, without requiring the reader to read 400 words of context to get to the answer. A section that opens with “To answer this question, we need to first understand the broader context of…” loses citation priority to a section that opens with “The answer is X, because Y and Z.”
Content freshness. For queries involving current data, recent events, or current best practices, recency significantly affects citation probability. Pages with 2025 or 2026 in headings, with dates visible in structured data, and with recent publication or update timestamps are preferred over technically-superior content from 2022.
Entity specificity. Content that references real, named entities — specific tools, companies, people, studies, statistics — is prioritised over generic advice. A
recommendation to “use a reputable analytics platform” loses to a recommendation that says “Google Looker Studio, pulling from BigQuery via a scheduled query, updates every 4 hours.”
Structural clarity. Q&A formatting, numbered lists, comparison tables, and clearly delineated sections are easier for AI systems to extract from than continuous prose. The AI needs to find a clean, extractable passage. Long paragraphs of narrative explanation require more inference to extract.
Authority signals. Domain authority matters, but not as a ceiling — low-DA sites with exceptional content get cited. What matters more is contextual authority: does the site consistently publish on this topic? Does it have structured data (Article schema, FAQ schema) that signals expertise? Is it cited in other authoritative content on the same subject?
Absence of conversion pressure. Pages that are heavily CTA-laden — where every other paragraph is pushing the reader toward a demo or purchase — receive lower citation rates than content-first pages. AI systems appear to prioritise pages that serve the informational purpose without significant commercial interruption.
The 9-Factor AEO Content Framework
Through 14 months of testing across 8 client content programmes, we’ve developed a framework for structuring content to maximise citation probability. We measure citation by tracking AI Overview appearances via Google Search Console and recording which client pages appear as sources.
Factor 1: Answer-First Structure
Every content section should lead with the direct answer to the query that section addresses. Not a teaser. Not context. The answer.
Before (standard editorial approach):
“When we think about conversion rate optimisation, it’s important to consider the relationship between landing page structure and user intent. Different traffic sources carry different intent signals, and the content a user expects to see varies considerably depending on how they arrived…”
After (AEO-structured approach):
“Traffic-source-specific landing pages improve conversion rates by 18-24% on average. A user arriving from Google search for a specific product term has different intent than a user arriving from a Meta brand awareness ad — serving both with the same landing page loses conversion rate from at least one of them.”
The after version can be extracted as a citation. The before version requires a reader to work through it.
Factor 2: FAQ Architecture
Explicitly structured Q&A sections are disproportionately cited in AI Overviews. The pattern is reliable enough that we now build FAQ sections into every content piece, structured with H3 headers phrased as direct questions and concise answers (80-150 words) immediately following.
FAQ sections should cover: the primary query the article targets, 3-5 related queries identified through People Also Ask and related search data, and common objections or misconceptions about the topic.
For example, an article about Smart Bidding should include FAQ sections not just for “how does Smart Bidding work” but for “does Smart Bidding work for small budgets,” “how long is the Smart Bidding learning phase,” and “can you use Target CPA and Target ROAS together” — all queries that appear in related search.
Factor 3: Proprietary Data and Original Research
AI systems cite original data more reliably than content that aggregates existing sources. A sentence like “across 41 accounts over 18 months, we found Smart Bidding produces CPA improvements of 26-34% under high-signal conditions” is more likely to be cited than “experts suggest Smart Bidding can improve CPA by 20-30%.”
This doesn’t require a formal research programme. It requires: keeping records of client results in a structured way, anonymising and aggregating those results into publishable benchmarks, and explicitly labelling internal data as proprietary (“our analysis of X accounts found…”).
Original data is harder for an AI to synthesise from multiple sources — it exists only in one place. That makes it citation-worthy.
Factor 4: Structured Comparison Content
Comparison queries (“X vs Y,” “best tools for Z,” “how X compares to Y”) generate strong citation demand, and comparison tables are extracted reliably. For every comparison article we write, we include at least one clearly structured table with named
tools, specific pricing, and specific capability differences.
The table format signals to the extraction system that comparable information is organised in a parseable structure. Prose comparisons are harder to extract cleanly.
Factor 5: Recency Signalling
For any content we want to compete for AI citation, we ensure: the year appears in the primary H1 and H2 headings where natural (“…in 2026”), the publication date is visible in article metadata, structured data (Article schema) includes a dateModified field updated when the content changes, and substantive updates happen at least annually for evergreen content.
Recency matters most for queries where information changes quickly: technology tools, pricing data, platform features, regulatory requirements. For more stable topics (foundational marketing principles, psychological frameworks), it matters less.
Factor 6: Entity Coverage
Named entities — specific tools, companies, platforms, studies, frameworks, people — make content more grounded and more extractable. We audit every pillar article for entity density before publication:
- At least 5 named tools mentioned with specific context
- Pricing data for tools mentioned where relevant (not generic “enterprise pricing”)
- Named studies or research sources with publication dates
- Real company examples (anonymised client data or named public case studies)
Content that is entirely generic — “use a good analytics platform” — has low entity density and low citation probability. Content that says “Northbeam’s algorithmic attribution model ($3,000-8,000/month for mid-market) uses Shapley value attribution to de-duplicate cross-channel conversion claims” has high entity density and significantly higher citation probability.
Factor 7: Expertise Signals
Citations cluster around content that demonstrates subject matter expertise rather than surface-level coverage. The signals AI systems (and Google’s quality raters behind them) use to evaluate expertise:
- First-person accounts of doing the work, not just describing it (“when we implemented this for a client…” vs. “the standard approach is…”)
- Acknowledgement of failure modes and exceptions (content that admits when something doesn’t work reads as more credible than content that only describes success)
- Technical specificity that wouldn’t be present in a surface-level treatment (specific error messages, exact configuration settings, precise API call syntax)
- Citations of non-obvious sources, not just the top-3 Google results everyone else links to
Factor 8: Content Completeness
For a given query, AI systems prefer pages that comprehensively cover the query intent rather than pages that partially address it. A query like “how to set up Google Enhanced Conversions” expects coverage of: what Enhanced Conversions is, when to use it, step-by-step setup, verification, and common implementation issues. A page that covers only the first two elements will lose citation to a page that covers all five.
We use competitor analysis for this: for every target query, we look at what the current top 5 ranking pages cover and identify topics they address that our content doesn’t. Closing those gaps is content completeness work.
Factor 9: Clean Page Architecture
Technical factors that affect AI extractability:
- Minimal JavaScript-dependent rendering: AI crawlers may not render JS-heavy pages fully
- Clean HTML structure: semantic headers (H1 -> H2 -> H3 hierarchy), not table-based layouts
- FAQ Schema markup: explicitly marks Q&A sections as such for AI systems
- Article Schema with accurate datePublished and dateModified
- Fast load time: pages that load within 2 seconds are crawled more completely
- No login walls or cookie consent interstitials that block crawler access to content
Tracking AEO Performance
Measuring AEO success requires different metrics than traditional SEO.
Google Search Console — AI Overview Impressions: GSC now reports AI Overview impressions separately under Performance > Search type > AI Overview. This shows you which queries are triggering AI Overviews, and whether your pages are appearing as citations. The click-through rate from AI Overview citations (typically 4-12%) is lower than standard organic, but the impressions represent brand exposure even when clicks don’t occur.
Perplexity Citation Tracking: Search Perplexity for your target queries. Note which sources it cites. This is a proxy for AI citation quality more broadly — Perplexity’s source selection reflects similar patterns to Google AI Overviews. If Perplexity cites your competitors but not you, your content structure is the variable to examine.
Direct URL checks in ChatGPT and Gemini: For pillar content, search the target query in GPT-4o and Gemini with browsing enabled. Does your site appear in cited sources? This isn’t statistically rigorous, but directionally useful for identifying whether your content structure is working.
Brand search volume as a leading indicator: Content that gets cited in AI Overviews creates brand awareness even when it doesn’t drive direct clicks. Track branded search volume in Google Trends and Search Console over time. A rising brand search trend in the absence of rising organic clicks suggests your content is being seen in AI contexts.
Semrush AI Overview tracker: Semrush’s position tracking now flags which target keywords trigger AI Overviews, and whether your site appears in those overviews. For accounts managing 50+ target keywords, this is the most efficient way to monitor AEO coverage at scale.
AEO in Practice: Content Architecture for a B2B SaaS Brand
Here’s how we structured an AEO-first content programme for a B2B SaaS client (HR software, 200-2,000 employee target market) over 8 months.
Starting point: 47 blog posts, primarily informational, organic traffic 12,400 monthly sessions, AI Overview citations: 3 pages cited across 8 queries.
Month 1-2: AEO audit of existing content. For each of the top 20 pages, we:
- Identified the primary query each page was targeting
- Searched that query in Google and noted whether an AI Overview appeared
- Checked whether the client’s page was cited in the overview
- Diagnosed why it wasn’t cited (answer-first structure, entity density, FAQ absence, freshness)
Of 20 pages audited: 14 had AI Overviews appearing; the client was cited in 2 of 14. Diagnosis across the other 12: 8 had answer-buried structure (buried after 300+ words of context), 7 lacked FAQ sections, 5 were >18 months old with no updates.
Month 2-4: Content restructuring (not rewriting from scratch — restructuring). For each of the 12 non-cited pages:
- Added direct-answer opening paragraphs to each major section
- Added FAQ sections (3-5 questions each, structured H3 questions with 100-word direct answers)
- Added or updated entity references (named tools with pricing, updated statistics)
- Updated publication dates and refreshed dateModified in Schema markup
Month 4-8: New content production, all built AEO-first from the outset. 16 new articles, each targeting a query with confirmed AI Overview presence, structured according to all 9 factors above.
Results at month 8: AI Overview citations increased from 3 pages across 8 queries to 19 pages across 31 queries. Organic traffic: 12,400 -> 17,900 sessions (+44%) despite AI Overviews absorbing click share on many queries (the net positive was from new citation-driven traffic and brand search growth). MQL pipeline attributed to content: +38%.
AEO for Perplexity, ChatGPT, and Gemini
Google AI Overviews are the largest AEO opportunity by traffic volume, but three other AI-driven answer systems are increasingly significant in how professionals research decisions:
Perplexity AI is the fastest-growing research tool among professionals in B2B contexts. Its citation model is notably different from Google’s: Perplexity almost always shows sources, its citations are more prominent, and it actively rewards content with original data and specific expertise. For B2B brands, Perplexity is disproportionately important relative to its absolute user numbers — it’s used by the exact buyer profiles (senior decision-makers doing pre-purchase research) that matter most.
Perplexity-specific optimisation: the same AEO principles apply, but directness matters even more. Perplexity’s interface shows the answer prominently, then sources. Content that answers the query in the first sentence, with supporting evidence in the following sentences, extracts cleanly. Long contextual introductions before the actual answer work against Perplexity citation.
ChatGPT with browsing (GPT-4o) is widely used for competitive research, vendor comparisons, and pre-purchase investigation. When browsing is enabled, GPT-4o retrieves and cites sources much as Perplexity does. Comparison content (“X vs Y”) and “best tools for Z” queries are the highest-value AEO targets for ChatGPT browsing, because these are the query types where buyer intent is highest and ChatGPT usage most common.
For comparison content, ensure your pages include: a clear recommendation or verdict (not just “both have pros and cons”), specific pricing data, and a named use case that distinguishes when to choose each option. ChatGPT users want a decision, not a balanced feature list.
Google Gemini in Search and in the Gemini app behaves similarly to AI Overviews in terms of source selection. The same structural and entity signals that improve AI Overview citation also improve Gemini citation. The primary difference: Gemini has stronger integration with Google’s knowledge graph, which means structured data (Schema markup) plays a larger role in source selection than in some other AI systems.
Practical multi-AI monitoring: weekly, search your 10 highest-priority target queries in Google (with AI Overviews), Perplexity, and ChatGPT. Note whether your content appears as a cited source. Maintain a simple tracking spreadsheet. Over 2-3 months, you’ll see which content is gaining multi-AI citation and which isn’t. The multi-cited pages are your AEO winners — understand what they have in common and replicate it.
Content Maintenance: The AEO Advantage of Freshness
One of the underappreciated aspects of AEO versus traditional SEO is how much more heavily freshness is weighted in AI citation selection.
In traditional SEO, a well-established page with strong backlinks can maintain its ranking for years with minimal updates. The authority accumulated over time offsets the lack of freshness. Google’s ranking algorithm gives significant credit to historical authority signals.
AI citation systems work differently. For any query involving current tools, current pricing, current best practices, or recent data, AI systems actively prefer recently updated sources. A page that was comprehensively written in 2023 but hasn’t been touched since will lose citation to a less authoritative page updated in 2026 that reflects current conditions.
This creates both a maintenance requirement and a competitive opportunity.
The maintenance requirement: every pillar article targeting queries with time-sensitive information (platform features, tool pricing, benchmark data, regulatory context) needs substantive annual updates — not just a date change, but actual content refresh with current data and current tool information.
The competitive opportunity: most competitors publish content and don’t update it. A systematic content maintenance programme that refreshes your top 20 pages every 12 months with current data will consistently outperform newer but unmaintained competitor content in AI citation selection.
Our practical process: for each pillar article, create a “freshness checklist” at publication: what are the time-sensitive elements (statistics, pricing, features, regulatory status)? Set a 12-month calendar reminder. When the reminder fires, update those specific elements before anything else. Add a new relevant case study or data point if one is available. Update the dateModified in the Article schema. This process takes 2-4 hours per article and meaningfully protects citation position.
Common AEO Mistakes
Trying to AEO-optimise every piece of content. AEO investment makes sense for queries where AI Overviews actually appear. For transactional queries (“buy X in Mumbai”), AI Overviews rarely appear. Standard SEO principles apply. Audit your target keyword set for AI Overview frequency before deciding where to invest AEO effort.
Optimising for citation volume without business purpose. Being cited in an AI Overview for “what is CPC” won’t meaningfully drive pipeline for a performance marketing agency. Target queries where the user who reads the AI Overview and then searches for more has a plausible path to becoming a client. AEO for brand-building queries and commercial investigation queries. Not for generic educational queries with no conversion pathway.
Treating AEO as purely technical. The content restructuring has to be accompanied by genuine content improvement. A page that leads with a direct answer but then gives incorrect or shallow information won’t sustain citation. The AI systems evaluate content quality, not just structure. The structure helps the AI find your answer; the quality
determines whether it cites you consistently.
Ignoring Perplexity, Gemini, and ChatGPT. Google AI Overviews are the biggest AEO opportunity by traffic volume, but Perplexity is growing rapidly in B2B research contexts. ChatGPT with browsing is used extensively for product research and comparison queries. An AEO strategy that focuses only on Google is leaving a growing share of AI-mediated discovery uncovered.
The Longer Game
AEO is not a campaign. It’s a structural shift in how you produce content.
The brands that will own AI-cited content over the next three years are the ones building content with genuinely original analysis, structuring it for extractability, and maintaining it with current data. The content farms producing generic, keyword-stuffed explainers will disappear from AI-cited results. They were already marginal in traditional search.
The bar AI systems set for citation-worthy content is, in a practical sense, the same bar readers set for actually useful content. The alignment is not accidental. The systems are optimised to surface what’s genuinely useful. Building for that bar is good strategy regardless of how the AI landscape evolves.
