Google AI Overviews vs ChatGPT vs Perplexity: AEO Differences

Direct Answer

Google AI Overviews, Google AI Mode, ChatGPT Search, and Perplexity are separate discovery surfaces, so a citation on one does not imply visibility on the others. They differ in interface, retrieval paths, source selection, prompt expansion, and how citations are displayed. But this does not require four unrelated content strategies. The shared foundation is still crawlable, useful, well-supported content and credible mentions. The platform-specific work is choosing which source ecosystems, formats, and measurement methods deserve priority.

In lesson 1.2 of the Ahrefs AEO course, Sam Oh reports that only seven of the top 50 cited domains appeared across Google AI Overviews, ChatGPT, and Perplexity in the course snapshot: 14% overlap. That is a useful warning against treating "AI search" as one leaderboard. It is not a permanent algorithmic rule. Source mixes can move quickly, and several of the leading domains had already changed by July 2026.

Watch the Ahrefs Lesson

Video and course credit: Sam Oh and Ahrefs. The official lesson page lists a 6:15 runtime. This article preserves the lesson's platform comparison, checks the cited studies against newer Ahrefs data, adds official guidance from Google, OpenAI, and Perplexity, and turns the findings into a repeatable measurement workflow. Ahrefs' datasets are observational snapshots, not disclosures of ranking algorithms.

Four Platforms, Four Different Surfaces

Google explicitly says there are no special AI-only technical requirements, files, or schema required for its AI features. OpenAI's publisher FAQ similarly says any public website can appear in ChatGPT search. Perplexity describes itself as an answer engine that searches the web, identifies sources, and synthesizes a response. These official descriptions matter because they put a limit on simplistic advice about "secret AEO factors."

What the Course Data Actually Shows

Query definition changes the result. In another Ahrefs comparison using roughly 3,000 shorter queries, Perplexity's exact-URL overlap with Google's top 10 was 65.07%, while ChatGPT's was 10%. For long-tail prompts, the same article reports 28.6% and 7.05%. These are not contradictory results; they show why a global platform average cannot replace a test of the prompts your customers actually use.

The Source Mix Had Already Changed by July 2026

The lesson says YouTube was AI Mode's leading cited domain. That was not the order in Ahrefs' July 2026 broad US snapshot: Reddit led AI Mode and YouTube ranked second. The current tables below use Ahrefs' reported mention share among top sources. They do not represent the percentage of every citation, and they should not be compared directly with the lesson's older 5.6% figure for YouTube URLs in a different AI Overview study.

Leaderboards are descriptive. A high share for Reddit does not mean a marketer should seed promotional comments. A high share for YouTube does not mean every topic needs a video. The useful question is whether that source type carries the evidence, audience, and format needed for a specific prompt. Google's guidance warns against inauthentic mentions and content created mainly to manipulate AI or search visibility.

Google AI Overviews: Search Fundamentals Still Apply

AI Overviews sit inside Google Search. Google's current documentation says the feature uses the Search index, core ranking systems, quality systems, Knowledge Graph, and other relevant signals. It can use query fan-out to search related subtopics and data sources concurrently. A page must be indexed and eligible to appear with a snippet, but it does not need special AEO markup.

Traditional rankings are still useful without being deterministic. Ahrefs' March 2026 update found that 37.1% of standard blue-link AI Overview citations ranked in the top 10 for the original query, 26.2% ranked from 11 to 100, and 36.7% did not rank in the top 100. Once all result blocks were included, 37.9% came from the top 10. This is a large change from Ahrefs' July 2025 result of 76.1%, and a reminder that fan-out and product changes can widen retrieval beyond the original result page.

Priorities for AI Overviews

• Keep the canonical page indexable, internally linked, fast, and eligible for a normal search snippet.
• Answer the question directly, then support it with dates, methodology, qualifications, and primary sources.
• Cover related decision needs coherently instead of generating a separate page for every synthetic fan-out.
• Use images and video when they genuinely explain the subject; keep important information available in text.
• Measure both original-query rankings and citation frequency because neither is a proxy for the other.

Google AI Mode: Similar Answer, Different Research Path

AI Mode is not merely a longer AI Overview. It supports follow-up exploration and typically produces a deeper response. Ahrefs' September 2025 study found that AI Mode answers were about four times longer and contained 2.5 times as many entities as AI Overviews. Yet their source overlap was only 13.7%, even when semantic similarity reached 86%.

The study also found that AI Mode cited Quora 3.5 times as often as AI Overviews and drew more heavily from some social sources. Those ratios describe one historical dataset. The durable lesson is that conversational depth creates more retrieval opportunities and a wider evidence graph, so AI Mode deserves its own tracking column.

Priorities for AI Mode

• Test multi-part, comparative prompts rather than only short keywords.
• Cover entities, alternatives, constraints, risks, implementation, and follow-up questions.
• Track whether the brand is named accurately even when an owned page is not cited.
• Compare AI Mode with AI Overviews using the same prompt, country, date, and clean-session protocol.

ChatGPT Search: Public Web Retrieval, Not a Publisher-Only Club

The course correctly observes that publishers and high-authority domains were prominent in its ChatGPT sample. Ahrefs found a median DR of 90 among the ranking pages in its September 2025 top-1,000 analysis, and OpenAI has licensing relationships with publishers. The overreach would be concluding that only licensed or DR 90 sites can appear.

OpenAI says any public website can appear in ChatGPT search. Sites should allow OAI-SearchBot if they want summaries, snippets, citations, and links. ChatGPT referral URLs include utm_source=chatgpt.com, which creates a practical analytics segment.

Ahrefs' May 2026 retrieval study adds another useful correction: 88.46% of the URLs that ultimately appeared as citations in its 1.4-million-prompt sample came through the general search reference type. That does not mean Google's top 10 controls ChatGPT. Exact URL overlap with Google remained low in other studies. It does mean searchable, relevant pages still matter alongside editorial mentions, licensing, Reddit, news, and other retrieval channels.

Priorities for ChatGPT Search

• Allow OAI-SearchBot while deciding separately whether to allow GPTBot for training.
• Create public pages that directly support the claims an answer should make about the brand.
• Earn independent editorial, community, review, and expert mentions because corroboration can matter even when an owned page is not cited.
• Track mentions, citations, factual framing, and utm_source=chatgpt.com referrals separately.
• Do not buy DR for its own sake. A current Ahrefs top-domain snapshot included specialist sites with materially lower DR than the earlier median.

Perplexity: Search Alignment Depends on the Prompt

Perplexity is explicitly search-oriented: it searches the web, identifies sources, synthesizes an answer, and shows citations. Its Pro Search can perform several searches across articles, academic papers, forums, videos, and other source types. This makes traditional search visibility a sensible starting point, but not a complete playbook.

Ahrefs' two overlap results show the range. Perplexity shared 65.07% of exact URLs with Google's top 10 in a short-tail query sample, but only 28.6% in a long-tail prompt sample. Domain overlap in the short-tail study was 80.58%. As prompts become more specific and fan out into supporting questions, the source set can move farther from the original SERP.

Priorities for Perplexity

• Start with pages that already rank for the core topic, then inspect citation gaps for longer decision prompts.
• Make primary evidence, definitions, comparisons, dates, and source references easy to extract.
• Test the modes and source types your audience actually uses; a research workflow is not the same as a quick answer.
• Validate that the visible citation supports the generated sentence, not merely the broad topic.

Five Claims That Need Qualification

Choose a Platform With Evidence, Not Hype

Market-wide numbers are useful context, but they answer different questions. Search-like prompt volume, referral traffic, monthly users, and assisted conversions are not interchangeable. Ahrefs estimated in February 2026 that ChatGPT handled search-like usage equivalent to about 11.86% of Google's daily search volume, while Google sent roughly 190 times more referral traffic. In another March 2026 sample of 74,752 sites, all AI assistants together supplied only 0.28% of total web traffic. An engine can influence a decision without producing a click, so the correct priority depends on the business.

Score each platform for one decision journey, not for the whole company. The highest total becomes the primary test surface for 30 days; the second becomes the comparison group. Revisit the choice quarterly because both audience behavior and source selection can change.

One Cross-Platform AEO Workflow

1. Build a fixed prompt set

Choose 10 to 20 prompts across awareness, comparison, risk, selection, and implementation. Include the buyer, country, constraints, and intended outcome. Do not rely on generic vanity prompts.

2. Establish controlled baselines

Run each prompt several times on the relevant surfaces. Record date, country, product, model when visible, login state, prior conversation, whether search occurred, brand mention, competitors, citations, and errors. AI answers are probabilistic, so a single screenshot is not a baseline.

3. Separate five outcomes

Track retrieval when observable, visible citation, unlinked brand mention, referral visit, and qualified conversion separately. A brand can influence the answer without receiving a click, and a cited page can receive traffic without generating a useful action.

4. Create a source matrix

For each platform and prompt cluster, list the cited domains, exact pages, source types, claims supported, freshness, and whether the source is owned, earned, community, video, data, or documentation. Look for recurring evidence needs, not just the biggest global domains.

5. Improve the shared foundation

Resolve crawl blocks, consolidate duplicate pages, add direct answers, identify entities clearly, cite primary evidence, show genuine update dates, expose limitations, and make key sections understandable when retrieved alone.

6. Add platform-specific distribution

For Google, strengthen Search eligibility and useful multimedia. For ChatGPT, allow OAI-SearchBot and earn relevant external corroboration. For Perplexity, make comparison-ready primary evidence discoverable. For AI Mode, expand the connected decision journey and test follow-ups.

7. Retest and connect to outcomes

Repeat the same protocol after recrawl and publication. Compare distributions, not one answer. Then connect changes to branded demand, referral sessions, assisted conversions, sales-call attribution, and corrected factual representation.

Use Brand Radar Without Letting the Tool Choose the Strategy

Sam demonstrates a blank search in Ahrefs Brand Radar, followed by the cited-domains report and a platform filter. That is useful for orientation. The stronger workflow starts with the brand, competitors, and a controlled prompt cluster.

1. Define the brand, products, competitors, country, language, and decision topic.
2. Compare mentions, cited domains, cited pages, and AI responses platform by platform.
3. Keep the date range and query set consistent before comparing source shares.
4. Find prompts where the brand is named but the claim is unsupported, outdated, or sourced to a weak page.
5. Identify domains that repeatedly support competitors but not the brand, then ask what genuine evidence is missing.
6. Export a monthly baseline and annotate launches, coverage, page updates, crawl changes, and major model releases.

A blank global report can reveal a platform's broad source mix, but it cannot tell a specialist company where to invest. Use the tool to observe; use customer research, business value, and source feasibility to decide.

Copy-Ready Platform Visibility Worksheet

Objective: compare AI visibility for one real customer decision.

Brand and canonical URL: [name / URL]
Audience, country, language: [specific segment]
Decision: [shortlist / purchase / implementation / risk review]
Platforms: [AIO / AI Mode / ChatGPT / Perplexity]
Test period: [dates]

Prompt set:
1. Awareness: [prompt]
2. Comparison: [prompt]
3. Fit and constraints: [prompt]
4. Risk and evidence: [prompt]
5. Implementation: [prompt]

Run each prompt five times per platform. Record:
- date, country, login state, conversation context, product/model
- live search visible: yes / no / unknown
- brand mentioned and exact framing
- competitors mentioned
- cited domains and exact URLs
- source type: owned / editorial / community / video / data / docs
- citation supports the claim: yes / partial / no
- material errors or missing qualifications
- referral session, assisted conversion, or self-reported attribution

Platform score:
- audience behavior: 0-5
- commercial relevance: 0-5
- current visibility gap: 0-5
- evidence and channel fit: 0-5
- technical eligibility: 0-5
- measurement quality: 0-5
- effort and policy risk: subtract 0-5

30-day action:
1. Fix one crawl or eligibility problem.
2. Improve one owned evidence page.
3. Earn or publish one useful third-party source.
4. Retest the fixed prompt set.
5. Report distribution changes and business outcomes.

Rules:
- Do not call one answer a ranking.
- Do not compare statistics with different denominators as a trend.
- Do not manufacture Reddit, review, or publisher mentions.
- Separate mentions, citations, clicks, and conversions.
- Label platform observations separately from causal claims.

Video Chapters

Bottom Line

Sam Oh's central lesson holds: optimizing for "AI search" as a single channel leaves blind spots. Google AI Overviews, AI Mode, ChatGPT, and Perplexity can answer similar questions with materially different sources. The right response is not four disconnected strategies or a race to imitate whichever domain leads this month's citation table.

Start with a strong, accessible evidence base. Measure a fixed set of customer-decision prompts across the platforms that matter. Learn which owned pages, publishers, communities, videos, and primary sources support the answers in your niche. Then prioritize the smallest set of changes that improves factual representation, credible citations, and qualified outcomes. Platform differences matter, but usefulness and evidence travel farther than any leaderboard.

Sources and Link Map

• Ahrefs video: How Ranking in Google AI Overviews, ChatGPT, and Perplexity Are Different - the embedded source lesson.
• Ahrefs Academy lesson 1.2 - official lesson page and runtime.
• Google Search Central: AI features and your website - official guidance on Search systems, fan-out, content, eligibility, and spam.
• OpenAI: Publishers and Developers FAQ - public-site eligibility, OAI-SearchBot, GPTBot, and referral tracking.
• Perplexity: What is an answer engine? - official description of web search, synthesis, and citations.
• Perplexity Pro Search - official description of multi-source research and follow-up.
• Ahrefs: AI Overviews vs AI Mode - 13.7% citation overlap, 86% semantic similarity, answer length, entities, and source differences.
• Ahrefs: ChatGPT and Perplexity citation overlap with Google - short-tail and long-tail exact-URL and domain comparisons.
• Ahrefs: updated AI Overview citation overlap - March 2026 top-10, positions 11-100, and outside-top-100 distribution.
• Ahrefs: ChatGPT's most-cited pages - September 2025 top-1,000 page sample, DR distribution, and traditional visibility gaps.
• Ahrefs: Why ChatGPT cites pages - May 2026 retrieval-channel study across 1.4 million prompts.
• Ahrefs: top domains in Google AI Overviews - July 2026 broad US snapshot.
• Ahrefs: top domains in Google AI Mode - July 2026 broad US snapshot.
• Ahrefs: top domains cited in ChatGPT - July 2026 broad US snapshot.
• Ahrefs: top domains cited in Perplexity - July 2026 broad US snapshot.
• Ahrefs: ChatGPT's estimated search-like volume versus Google - February 2026 volume and referral context.
• JQ AI SYSTEMS: How AI Search Engines Work - retrieval, query fan-out, citations, and a source-readiness checklist.
• JQ AI SYSTEMS: What Is AEO? - terminology, course map, and a 30-day baseline.