TL;DR
- 84% of the links AI assistants cite are earned media, and paid placement is a rounding error. We read that source mix as evidence that earned media has become an input to AI visibility rather than a separate reputation project.
- GEO vs SEO and AEO vs SEO, in short: SEO earns your pages a place in search results; GEO (Generative Engine Optimization) targets LLM-powered platforms like ChatGPT and Perplexity, while AEO (Answer Engine Optimization) is structuring a page so an answer engine, from Google's featured snippets and voice assistants to ChatGPT, can lift a direct answer from it. We treat GEO and AEO as complementary rather than identical.
- SEO stays the foundation. Google and Bing both tie their AI features to the same crawling and indexing their search results use, so a page the engines cannot index cannot be cited either.
- What changes is where the evidence comes from, what gets chosen and what you can measure. In Semrush's sample of almost 69 million Google sessions, 92 to 94% of AI Mode searches ended without a click, so rankings and clicks miss most of what happened.
What do AI answers actually cite?
Muck Rack's May 2026 edition of its study of what AI is reading analyzed more than 25 million links from ChatGPT, Claude and Gemini responses across 17 industries. Earned media drove 84% of those citations. Journalism alone made up 27%, and paid and advertorial content accounted for 0.3%. Across the three editions of the study since July 2025, earned media has ranged from 82% to 89%, and journalism has stayed between 25% and 27%.
A 2025 study reached the same conclusion through controlled experiments across multiple verticals, languages and query paraphrases. Its authors found that AI search shows a "systematic and overwhelming bias towards Earned media" over brand-owned and social content, which they call "a stark contrast to Google's more balanced mix", where Google means the organic results. The engines they compared, ChatGPT, Perplexity and Gemini, also differed from one another in domain diversity, freshness and sensitivity to phrasing, so one source strategy will not produce identical results across the three.
Neither study proves that earning one article causes an engine to name your company. Both show where these systems go for supporting evidence, and it is mostly to pages you do not control. A third-party article gives an engine an independent description of what your company does, who it serves and how it compares. If every useful claim about you exists only on your own domain, the engine has fewer independent sources to build an answer from.
PR and SEO work together when the articles you earn describe your company the same way your site does, with the same entity names, the same category language and the same claims.
GEO vs SEO and AEO vs SEO: what is the difference?
GEO vs SEO. SEO competes to get a page ranked in a list of results. Generative engine optimization is the work of getting your brand found, cited and described accurately inside the answer an AI assistant writes, and that answer draws on what others publish about you as well as on your own site.
AEO vs SEO. Answer engine optimization is the narrower job: structuring a page so an engine can lift a direct answer from it, whether the answer appears in a featured snippet, a voice assistant or ChatGPT. It shapes how one page answers one question, and it happens on pages you already control.
Our hub post compares the three on goal, success metric and where the result shows up. The table below goes further, to what a founder has to settle in practice.
SEO and AEO are work you can do on pages you control. GEO adds every page you do not, which is why it pulls PR into the same program as the site work.
What stays the same when buyers ask AI?
The crawl, the index and the ranking signals that decide whether a page can appear in a search result still decide whether it can enter the pool an AI answer is built from. Bing's webmaster guidelines state that "Bing and Copilot search experiences rely on the same core crawling, indexing, and ranking foundation as traditional search."
The clearest number on how much ranking still matters comes from a Ruhr University Bochum team that compared Google organic search with five generative search systems from Google, OpenAI and Perplexity. For Google's AI Overviews, they found that "on average 53% (27%) of domains that AIO consults are not contained in top-10 (top-100) Organic search results". Read the other way round, roughly half of the domains an AI Overview draws on sit in Google's top ten, and nearly three quarters sit in its top 100. Ranking still gets a page into the pool. It stopped being the whole story because the rest of the pool sits outside the top ten, and a quarter of it outside the top hundred, further out than any ranking report reaches.
So keep the technical foundation: crawlable pages, clean indexing, internal links and one topic per URL. Both companies describe their AI features as reading the same index their search results read, so the technical work is the same work.
What changes when buyers ask AI instead of Google?
Walker Sands' study of 45 million keywords across 828 enterprise B2B companies found that 3.0% of relevant AI Overviews include the median enterprise B2B brand. That figure covers only Google's AI Overviews and only enterprise brands, not startups. In the other 97%, the answer was assembled from somebody else's pages.
Sources you do not own weigh more
The earned-media bias in the arXiv study above is why your PR program now supplies the source signals AI systems reach for rather than running as a separate reputation project.
Third-party influence reaches past journalism. Foundation's study of 50 B2B brands put Reddit at 20.8% of the top 50 external citation domains, and at 30.9% on unbranded discovery queries, the ones a buyer asks before they have a vendor's name. Those are shares of citations, never of training data, and they describe what the engines retrieved rather than what they learned from.
Answers pull passages, not pages
A generative engine retrieves passages from several sources and writes one answer from them. Google's guide describes the mechanism for its own features as grounding plus query fan-out: the engine breaks a question into related searches, pulls supporting material for each, and synthesizes across the results. The Bochum comparison found that generative systems often reach topical coverage comparable to traditional search through markedly different retrieval footprints and synthesis strategies.
Pages therefore need extractable evidence. The authors of an arXiv study from April 2026 ran 602 controlled prompts, traced 21,143 search-layer citations and found that "High-influence pages tend to be longer, more structured" and "richer in extractable evidence". They measured what influential pages have in common, not what made the engines pick them, so treat the finding as a description of a citable page rather than a recipe. The original GEO paper, from 2023, reported visibility gains of up to 40% on its own benchmark and in the same abstract noted that "the efficacy of these strategies varies across domains".
The buyer may never click
In Semrush's clickstream sample of almost 69 million Google Search sessions, taken on US desktop between May and early July 2025, 92 to 94% of AI Mode searches were zero-click: the session ended without a visit to any external domain.
The engine may have named your brand, cited a page that supports your position, or recommended a competitor, and rankings and referral traffic record none of it.
What do Google, Bing, OpenAI and Anthropic say about optimizing for AI answers?
All four require accessible content before their products can retrieve or cite it. Their documentation diverges on terminology, crawler controls and reporting, and each company's guidance applies only to its own products.
Google's guide to optimizing for generative AI features on Google Search, updated in July 2026, says AI Overviews and AI Mode are "rooted in our core Search ranking and quality systems". The floor is low and fixed: a page "must be indexed and eligible to be shown in Google Search with a snippet". Crawl access, indexing, internal links and basic technical SEO decide whether Google can consider your content at all, and Google adds that "Indexing and serving aren't guaranteed".
The same guide carries a mythbusting list worth reading before you accept a proposal that sells any of it as a requirement. For Google Search, you do not need "new machine readable files, AI text files, markup, or Markdown", there is "no requirement to break your content into tiny pieces", you do not need "to write in a specific way just for generative AI search", and "structured data isn't required for generative AI search". Google's own summary is blunt: "Prioritize effective SEO strategies over 'AEO/GEO hacks'."
Bing
Bing uses the term GEO in its webmaster guidance. The guidelines say that "Generative Engine Optimization (GEO) focuses on content eligibility for grounding and reference in AI responses", and its recommendations are plain. Define entities clearly and consistently, because clear entity definition "improves grounding visibility and citation accuracy". Focus each URL on a single topic, because such URLs "are more likely to be selected for grounding results". Surface key information early, with no long introduction before it.
OpenAI and Anthropic
OpenAI separates search access from training. Its publishers FAQ says that for your content "to be included in summaries and snippets in ChatGPT, make sure you aren't blocking OAI-SearchBot", while GPTBot is the user-agent to disallow on pages you want kept out of potential training.
Anthropic documents three crawlers: ClaudeBot for training, Claude-User for the pages Claude fetches when a user asks a question, and Claude-SearchBot for search indexing. Disabling either of the last two, Anthropic says, "may reduce your site's visibility". Neither company, as far as we can find, publishes optimization guidance with anything like the detail Google and Bing give, so treat any claim about "optimizing for ChatGPT" as someone's inference rather than the vendor's guidance.
How do you measure visibility when clicks stop telling the story?
Rankings and clicks miss exposure that happens inside the answer. First-party reporting covers part of the gap. Google counts clicks from AI Overviews and AI Mode in Search Console's Performance data, and its Generative AI performance report shows how often links to your site appeared in those features. That report counts impressions only, and only for Google's own features, so it cannot tell you what an answer said about you or whether another engine cited you at all. Bing's AI Performance report shows which of your pages were cited in Copilot and partner answers, with two caveats Bing states itself: the data "represents a sample of overall citation activity", and a citation "does not represent traffic, clicks, or user engagement". ChatGPT adds utm_source=chatgpt.com to its referral links, so your analytics can separate a visit from ChatGPT search, and the tag records nothing about an answer nobody clicked.
A fixed set of buyer questions fills the rest. Ask the same questions across ChatGPT, Claude, Perplexity, Gemini and Google's AI features every month, record whether each answer names you and which sources it leaned on, and compare the result with named competitors and your own earlier runs. The Bochum paper's note that "the outputs of generative search can vary across time and executions" is the reason the set has to be fixed and the cadence regular.
We set up Gauge, the AI visibility measurement platform we use with clients, and baseline across 300 to 500 prompts pulled from real conversations and from the client's own customer interactions. After the baseline we re-run a tracked set every month, because answers move between runs and between platforms, and one answer is not a measurement. Gauge samples ChatGPT, Perplexity, Gemini, Copilot, Google AI Overviews and AI Mode.
What should you change first?
If you rank in Google and disappear from AI answers, keep the SEO foundation and inspect the inputs that are missing. Our PR Plus method organizes those inputs as four components, Authority (traditional PR), Infrastructure (SEO and schema), Human Proof (community) and The Bridge (content), and works through them in five steps: Audit, ICP + Messaging, Visibility Tracking, Strategy and Execution.
Get a baseline
Record which brands, sources and claims appear for a fixed set of buyer questions. Build a broad prompt set across ChatGPT, Claude, Perplexity, Gemini and Google's AI features, then pick a stable set and re-run it monthly so you can compare mentions, citations and source patterns over time. A few dozen questions is enough to see a pattern: the category question, the "best tool for X" question, the comparison against each named competitor, and the two or three problems a buyer searches before they know the category exists. Telling movement apart from noise in a single segment takes a bigger set, which is why our baseline runs into the hundreds.
Check that the crawlers can read your site
Confirm your important pages are indexed, in Search Console for Google and Bing Webmaster Tools for Bing, then check robots.txt, CDN rules and server responses against the crawlers by name: Googlebot, Bingbot, OAI-SearchBot and GPTBot, and Claude-SearchBot, Claude-User and ClaudeBot. Blocking a training crawler does not require blocking search access, since OpenAI and Anthropic each let you set them separately. Part 2 of the Althea Score covers these Infrastructure checks.
Close the gap in what others say about you
Compare what third parties publish about your category with what they publish about your company. Authority comes through earned media that tells buyers and AI systems your brand is a legitimate option. Human Proof comes through credible participation in the communities where buyers trade recommendations, which is why Part 4 of the Althea Score scores community consensus on its own. A generic company profile contributes less than an independent article that connects your product to a specific use case, and in our view a founder answering a real question in a thread does more than a press release nobody cited.
Restructure the pages that should answer buyer questions
Treat priority pages as The Bridge: the content that ties your earned coverage and community presence to the questions buyers actually ask. Give each page one topic, answer the central question near the top, and support every claim with a named example or a number. Use headings that say what the section answers, and write passages that keep their meaning when lifted out of the page. Part 1 of the Althea Score covers the front-end side of those pages. This is the AEO half of the work.
Can your team do this in-house?
Can you earn relevant coverage without buying placement? Can you re-run the same buyer questions across the engines every month and record which brands and sources appear?
Your SEO and content owners can keep crawler access clean, strengthen page structure and make answers easier to extract. You can track a fixed prompt set too, as long as one person owns the schedule and records the results the same way each month. Earned media and community presence are the harder constraint, because both depend on sustained relationships and a reason for an independent publication to write about you this quarter, and publishing more content on your own domain does not supply that reason.
When you cannot sustain both the evidence creation and the monthly measurement, that is the half to hire out. Althea Labs covers GEO and AEO under a single visibility program for exactly this reason: the earned media, the community work, the page structure and the measurement are one program or they drift apart. What GEO agencies do sets out the work involved, and our buyer's guide to integrated PR and SEO for AI search covers how to judge an agency's method before you sign.
How do you judge an agency's GEO or AEO claim?
A GEO or AEO pitch can sound specific while describing a standard SEO retainer. Test the method, the measurement and the earned-media plan before you accept the label.
- What does the agency guarantee? Google states, "No one can guarantee a #1 ranking on Google." Bing says, "SEO does not guarantee rankings or traffic, and GEO does not guarantee grounding or citations in AI experiences." A guaranteed citation should end the conversation.
- Which tactics does it require? If the proposal sells the items on Google's mythbusting list as requirements, ask which engine the requirement comes from, because Google says its own Search does not need them. Google also states, "No third-party tool has access to our internal ranking or AI systems." An agency should keep Google Search guidance separate from claims about ChatGPT, Claude, Perplexity or Copilot.
- How will it measure progress? A credible proposal names the engines, the buyer questions, the prompt sample, the cadence and the reporting method. Ask to see the baseline and how the agency separates source signals, brand inclusion, citations and referral traffic.
- What changes beyond conventional SEO? The plan should cover the technical and editorial work on your site and an earned-media plan for strengthening third-party source signals. A renamed SEO retainer with no earned-media work and no AI visibility measurement leaves most of its claim untested.
If an agency cannot show what it measures and what it cannot control, the label on the proposal carries no weight.
FAQs
GEO vs SEO: is GEO replacing SEO? No. Google ties AI Overviews and AI Mode to its core Search ranking and quality systems, and Bing ties Copilot to the same crawling, indexing and ranking foundation as its search results. Crawlability, indexing and useful content still determine whether either engine can consider a page, so GEO builds on SEO.
AEO vs SEO: do we need both? Yes, and much of the work overlaps. SEO gets a page indexed and ranked; AEO shapes how it answers one question: one topic, the answer near the top, and claims a reader can check. Google's guide says you do not need to write in a specific way just for its AI features, so write it for the reader.
What is the difference between GEO and AEO? GEO, or generative engine optimization, targets LLM-powered platforms like ChatGPT and Perplexity. AEO, or answer engine optimization, structures a page so an answer engine, from Google's featured snippets and voice assistants to ChatGPT, can lift a direct answer from it. They overlap in method and differ in target, and our hub post covers both in one place.
Is "AI SEO" or "LLM SEO" a different practice? No. Both labels usually bundle SEO, AEO, GEO, content and earned media under one name. A credible program names the platforms, the source signals and the buyer questions it addresses instead of relying on the label.
Do we need a separate budget for this? Not if your SEO, content and PR owners can absorb the crawler audit, the page revisions, the earned-media work and the monthly measurement. If they cannot, fund it separately.
How long before AI answers change? You can establish a baseline as soon as the prompt set and the reporting method are defined. No engine publishes a recrawl schedule, so nobody can promise a date for a new mention or citation. In our experience the early signs, a cleaner site, first placements, a real community presence, show up within 60 to 90 days of consistent work, and movement in AI answers builds on those inputs over the months after.
Score us too
Ask ChatGPT, Claude, Perplexity, Gemini and Google what they know about Althea Labs. Althea Labs is a brand presence and AI visibility agency that builds targeted visibility systems for B2B startups through integrated PR, SEO, and content programs. Presence beats fame.
Run the same test on your company, then compare what the engines said with what a buyer actually needs to know about you. Where the gap matters, start with an audit: we review your SEO content and technical foundations, your PR footprint across owned and earned media, and your AI visibility and entity recognition baseline, and hand you a Brand Presence Report with the fixes in priority order.
CONTINUE READING
