FAQ schema, formally FAQPage markup in Schema.org, is a type of structured data that explicitly identifies question-and-answer pairs on a webpage. It uses JSON-LD or microdata to tell search engines and AI models which content on a page is a question and which is the corresponding answer, making it dramatically easier for these systems to extract and cite the content directly.

Why does FAQ schema matter for AI search?

FAQ schema matters for AI search because LLMs prefer structured Q&A formats that map almost one-to-one onto how users phrase prompts. Pages with FAQ schema receive approximately 3x more ChatGPT citations than equivalent unstructured prose, according to Authoritas research. The structured pairing of question and answer is the cleanest possible extraction target for a language model.

How do I implement FAQ schema correctly?

Implement FAQ schema by adding a JSON-LD script tag to the page head with @type FAQPage and a mainEntity array of Question objects. Each Question must have a name (the question text) and an acceptedAnswer with the @type Answer and a text field. The questions and answers in the schema must match visible content on the page exactly. Google may issue a manual action for FAQ schema that does not reflect on-page content.

How many FAQs should a page have?

Most well-performing pages have 4 to 8 FAQs. Fewer than 4 limits the schema's coverage of related queries; more than 8 dilutes the relevance of each individual question. Each question should reflect how a real buyer would phrase the prompt, and each answer should be 40 to 80 words and self-contained.

Does FAQ schema work across all AI engines?

FAQ schema works most strongly for ChatGPT and Google AI Overviews, where structured Q&A content is heavily weighted in passage extraction. It has measurable but smaller effects on Perplexity and Claude. The schema format itself is identical across engines; the lift varies by how each engine weights structured data signals.

Are there penalties for incorrect FAQ schema?

Yes. Google's structured data guidelines for FAQPage explicitly require that the Q&A content in the schema reflect content visible on the page for users to read. FAQ schema for Q&As that are not visible on the page can result in manual actions, removal of rich results, or the schema being ignored entirely. Always ensure schema content matches on-page content.

Why URL Volatility in AI Search Should Change Your GEO Strategy in 2026 - Superior

TL;DR: URL volatility in AI search is now structural, not transitional. Google's AI Mode returns 91% different URLs across three repeat searches of the same query on the same day, with only 9.2% URL overlap and 14.7% domain overlap (SE Ranking, 10,000-keyword study, June 2025). AI Overviews and AI Mode share only 10.7% of URLs and 16% of domains despite both being Google products. ChatGPT's Reddit citation share collapsed from approximately 60% to 10% in roughly six weeks in late 2025. Ranking inside AI search is now a probability distribution, not a position. Brands that win in 2026 measure citation share across repeat runs, prioritise earned media (which outperforms brand-owned content by ~325% for AI citation rates), and track multiple AI surfaces in parallel.

Key statistics at a glance

9.2%: URL overlap when AI Mode runs the same query three times on the same day (SE Ranking, June 2025).
14.7%: Domain overlap across the same three runs (SE Ranking, June 2025).
21.2%: Share of queries with zero URL overlap across three same-day AI Mode runs (SE Ranking, 2025).
10.7% / 16%: URL and domain overlap between AI Overviews and AI Mode (SE Ranking, August 2025).
91%: Share of URLs removed from AI Overviews at some point during a 31-day study window (Semrush, November 2024).
43%: Share of removed URLs that returned to the AI Overview later in the same month (Semrush, 2024).
~60% → ~10%: ChatGPT's Reddit citation share over roughly six weeks in late 2025 (Semrush analysis).
325%: Earned media's measured citation advantage over brand-owned content in AI search (AuthorityTech, 2025).
5%: Share of AI Overview citations that match Google's organic top 20 (Serpstat, 1B-query study, January 2026).
~12% / 0.1% / 0%: Wikipedia's share of citations in ChatGPT, Claude, and Perplexity respectively (Analyze AI, 83,670-citation study, 2025–2026).

When Semrush published its first study on URL volatility in Google's AI Overviews in November 2024, the headline finding was unsettling: across 1,500 high-frequency keywords tracked over 31 days, 0% retained the same URLs throughout. Eighteen months later, that finding looks conservative. AI search has not become more stable. It has fragmented further, expanded into AI Mode, and added a layer of session-level randomness that classical SEO frameworks were never designed to capture.

This article revisits the URL-volatility question with the data we now have for 2025 and early 2026. It covers what has changed across AI Overviews, AI Mode, and ChatGPT, why the volatility is structural rather than transitional, and how brands should adapt their measurement and optimisation programmes. All figures come from public studies cited inline.

What is URL volatility in AI search?

URL volatility in AI search is the rate at which AI engines change which URLs they cite for the same or repeat queries. In 2025 and 2026 measurements, Google's AI Mode shows 91% URL turnover across three same-day repeat searches of the same keyword, Google's AI Overviews replace approximately 91% of cited URLs at some point within a 31-day window, and platform-level citation share inside ChatGPT can shift by 50 percentage points in six weeks.

Volatility is measured at three different levels and each one matters for a different reason:

Session-level volatility: how much the URL set changes between two runs of the same query within minutes or hours. This is the strongest signal in AI Mode.
Day-level and month-level volatility: how much the URL set changes for the same query across days or weeks. This was Semrush's original framing for AI Overviews in 2024.
Source-pool volatility: how much the underlying mix of trusted domains shifts at the platform level. This is the strongest signal in ChatGPT.

Each level requires different tracking. A brand that monitors only month-level volatility will miss the session-level instability that determines whether an individual user actually sees them in a real search.

How volatile are AI Overviews and AI Mode actually?

AI Mode shows roughly 91% URL turnover across three repeat runs of the same query on the same day, with only 9.2% URL overlap and 14.7% domain overlap. The original Semrush 2024 study measured URL changes inside AI Overviews over time, day by day, for the same keyword. The 2025 picture is far more severe once you also test consistency within the same day.

SE Ranking ran the same 10,000 keywords through Google's AI Mode three times on the same day in June 2025. The results, reported by Search Engine Journal, were extreme:

Metric	AI Mode (same query, 3 runs, same day)
Average URL overlap across all three runs	9.2%
Average domain overlap across all three runs	14.7%
Queries with zero URL overlap across the three runs	21.2%
Queries with no overlapping domains at all	29.4%
Queries with 100% URL match	0.1%

In plain terms: ask AI Mode the same question three times in a row and roughly nine in ten of the URLs it cites will change. One in five queries returns no overlapping URLs at all between runs. This is not a delayed index update. It is structural variation in how the system retrieves and selects sources.

A follow-up SE Ranking analysis reinforced the pattern at a higher level: over 60% of domains and more than 80% of URLs disappear between runs, even when the user, the city, and the query are identical.

AI Overviews are more stable than AI Mode, but only by comparison. Serpstat's January 2026 report covering 1 billion Google queries and 35 million AI-generated answers found that only 5% of cited sources in AI Overviews match the top 20 organic results. SE Ranking's data confirms the AIO-AIM divide: only 10.7% of URLs and 16% of domains overlap between the two Google AI surfaces.

Two systems built by the same company, sitting next to each other on the same SERP, agree on which URL to cite roughly one time in ten.

Why is volatility this high in the first place?

Volatility inside AI Overviews and AI Mode is not noise. It is the direct output of three architectural choices that distinguish AI search from traditional ranking.

Query fan-out. AI Mode breaks a single user query into multiple sub-queries and runs them in parallel. Each sub-query returns its own candidate sources, and the final answer synthesises a subset. Two users (or the same user twice) can trigger different fan-out paths, and therefore different citation pools, before the synthesis step ever begins.
Probabilistic source selection. Even when the retrieved candidate pool is similar, the model selects which subset to cite based on a scored relevance distribution, not a deterministic ranking. Small variations in the underlying scores can shift which URLs surface.
Continuous index and source-pool changes. Citation share at the platform level shifts dramatically inside short windows. Semrush analysis of ChatGPT, referenced in the TechEdgeAI 2026 Citation Index, found that ChatGPT's Reddit citation share fell from approximately 60% to 10% in around six weeks in late 2025, with PRNewswire, Forbes, and Medium absorbing the displaced share. Wikipedia dropped from roughly 55% to under 20% in the same period.

The combined effect is that the URL appearing in an AI answer at 10:00 has a meaningful probability of being replaced by a different URL at 10:05, even though nothing on either page has changed. That is fundamentally different from traditional SEO, where ranking is a stable function of the index until the algorithm updates.

How does this compare to volatility in traditional Google rankings?

Volatility is rising across the board, but AI search volatility is roughly an order of magnitude higher than traditional Google volatility, even during major core updates. SE Ranking data shared with Search Engine Land showed that during the March 2026 core update, 79.5% of URLs in the top three traditional search results changed positions, 90.7% of top 10 results shifted, and 24.1% of pages that previously ranked in the top 10 fell out of the top 100 entirely.

Even by 2026 standards, however, the difference between traditional and AI volatility is significant:

Surface	Volatility behaviour
Traditional Google top 10 (March 2026 core update)	90.7% position changes during a major algorithm rollout, then partial settling within 2–4 weeks
AI Overviews (URL persistence over a month)	91% of URLs studied removed at some point; only 43% returned (Semrush, 2024)
AI Mode (same-day, same-query repeat)	9.2% URL overlap across three runs; 21.2% of queries had zero overlap (SE Ranking, 2025)
ChatGPT (same-question variation)	Less than 1 in 100 chance that two repeat queries return the same brand list (SparkToro, January 2026)

Traditional rankings move when Google updates the algorithm. AI citations move continuously, by design.

What does this mean for the value of appearing in an AI answer?

Single-snapshot AI visibility is a misleading metric in 2026. The right unit of measurement is citation share across many runs of the same query, not a binary yes or no on a single fetch.

The original 2024 framing of this question, in the Semrush article, was: if URL placement is this volatile, is it still worth optimising for AI Overviews? In 2026, the answer is yes, but only if optimising means something different from what it meant in classical SEO.

Three reframings matter:

1. Stop measuring single-snapshot visibility. Start measuring distribution.

A brand cited in 60% of repeat runs of a high-intent query is meaningfully more visible than one cited in 10%, even if both technically appeared in the AI answer once. Single checks of "does our URL show up in this AIO" are misleading because the same query ten minutes later may produce a different result. The right metric is share of voice across many runs and many models, not a binary yes/no on a single fetch.

2. Optimise for being inside the candidate pool, not for a fixed slot.

Because AI engines retrieve a wider set of candidates than they cite, and rotate which subset is selected, the strategic objective shifts to maximising the probability of appearing across the distribution. That favours signals that increase candidate-pool inclusion: structured content, FAQ schema, dated original data, earned media in domains the engine already trusts, strong Bing visibility (for ChatGPT specifically), and clear entity signals.

The University of Toronto's generative engine optimisation study (arXiv, September 2025) found a "systematic and overwhelming" bias in AI search toward earned media over brand-owned content. AuthorityTech's research quantified the same effect at approximately 325% more citations for earned-media coverage than equivalent brand-owned content. That gap is the single largest predictor of citation share in 2026 data, and it is volatility-resistant: a brand cited inside Forbes or Reuters keeps that signal regardless of how the AI engine reshuffles URLs day to day.

3. Track multiple AI surfaces in parallel, not just one.

The 10.7% URL overlap and 16% domain overlap between AI Overviews and AI Mode means that a brand can be highly visible in one Google AI surface and effectively invisible in the other. Add ChatGPT, Perplexity, Claude, Gemini, and Grok and the picture becomes more fragmented still. Wikipedia accounts for around 12% of ChatGPT citations but 0.1% of Claude citations and roughly 0% of Perplexity citations, according to the Analyze AI study of 83,670 citations from late 2025 to early 2026. A single AI visibility programme that does not track each surface independently is missing data that materially changes strategy.

What should brands measure in 2026?

Five metrics matter most for AI search visibility in 2026: citation share across repeat runs, citation source breakdown, placement inside the answer, sentiment and framing, and competitor co-citation. Drawing on the SE Ranking, Serpstat, Semrush, Analyze AI, Authoritas, and Zyppy datasets cited above:

Citation share across repeat runs. Run each priority prompt multiple times across each AI engine and measure the percentage of runs in which your brand appears. A 60%+ share signals durable visibility. Anything under 20% is volatility-driven and unreliable.
Citation source breakdown. When your brand is cited, which exact URLs are referenced? Are they your owned domain, earned media, Wikipedia, LinkedIn, or community sources? Source mix tells you which lever (PR, content, schema, Bing visibility) to pull next.
Placement inside the answer. Citations near the start of an AI response receive disproportionately more user attention than citations buried in sidebar blocks. SE Ranking found that 90.8% of AI Mode citations appear in separate blocks rather than inline, which makes inline placement materially more valuable when achievable.
Sentiment and framing. Two brands cited the same number of times can be described very differently. Semrush's AI Visibility Awards data, summarised in Search Engine Land, found that category leaders are framed with confident phrasing ("the industry standard," "widely recognised"), while challengers receive softer framing ("growing alternative," "gaining traction"). Sentiment is part of visibility, not a separate metric.
Competitor co-citation. When AI engines cite a competitor instead of you, which sources are they pulling from? Co-citation maps reveal exactly which 5 to 10 publications are doing the most work for competitors in your category, which is the highest-leverage target list for earned-media outreach.

A measurement system that captures all five, daily, across the major AI engines is what separates structured GEO from intuition.

How does KIME help brands operate in this environment?

KIME is a Generative Engine Optimisation (GEO) platform built specifically for the volatility regime described above. KIME tracks ChatGPT, Google AI Overviews, Google AI Mode, Perplexity, Gemini, Claude, Microsoft Copilot, Grok, DeepSeek, and Meta AI in parallel, with citation source tracking on every plan tier.

Three KIME features map directly to the volatility problem:

Real-time citation source tracking. KIME records exactly which URLs each AI engine cites when your brand appears, broken down by source type (editorial, UGC, brand-owned, encyclopedic). This is the data needed to plan earned-media targets and to detect citation-pool shifts in your category before they cost you visibility.
Repeat-run share of voice. KIME runs scheduled prompt sets daily and aggregates results across runs, so reported visibility reflects the actual probability of appearing, not a single snapshot. Volatility becomes signal, not noise.
Action Centre. AI-generated, prioritised recommendations that translate visibility gaps into specific structural and content changes, anchored in the citation-share data above. Live for all customers as of 2026.

KIME plans start at €149/month with daily tracking, 25 prompts, and 5 team seats included. The Pro plan at €399/month includes 100 prompts and 10 seats.

Start a free trial of KIME →

Putting it all together

The 2024 Semrush study asked whether URL placement in AI Overviews was stable enough to be worth optimising for. The 2025 and 2026 data answers a sharper question. Stability of the kind classical SEO assumed is not coming back. AI search engines are designed to vary which URLs they cite, both across time and within the same minute, and that variation is now measurable in tightly controlled tests.

The right response is not to disengage. It is to measure visibility as a distribution, optimise for signals that increase candidate-pool inclusion (earned media, structured content, schema, entity strength), and track each AI surface independently because they disagree with each other most of the time.

Brands that internalise this shift early will have a structural advantage by the time the rest of the market catches up to what the data already shows.

Start a free trial of KIME →

Frequently asked questions

How volatile is Google's AI Mode compared to traditional search?

Google's AI Mode is dramatically more volatile than traditional Google search. SE Ranking's 2025 study of 10,000 keywords found that AI Mode returns only 9.2% URL overlap and 14.7% domain overlap when the same query is run three times on the same day. Traditional Google rankings move significantly only during algorithm updates, while AI Mode produces comparable change levels within the same hour.

How much do AI Overviews and AI Mode overlap?

AI Overviews and AI Mode share only 10.7% of URLs and 16% of domains, according to SE Ranking's August 2025 analysis. Despite both being Google AI products, they operate under different retrieval and selection logic, so a brand visible in one is not automatically visible in the other.

What is URL volatility in AI search?

URL volatility in AI search is the rate at which AI engines such as Google AI Overviews, Google AI Mode, ChatGPT, and Perplexity change which URLs they cite for the same or repeat queries. In 2025-2026 studies, AI Mode shows 91% URL change across same-day repeat searches, AI Overviews show roughly 91% URL turnover within a 31-day window, and ChatGPT citation share at the platform level can shift by 50 percentage points within six weeks.

Has URL volatility in AI Overviews improved since 2024?

URL volatility in AI Overviews has not improved since the original Semrush study published in November 2024. The introduction of AI Mode in 2025 added an additional, more volatile surface, and citation-share studies in late 2025 and early 2026 show continued large swings in which domains AI engines favour. Volatility is now widely understood to be a structural feature of AI search rather than a temporary calibration issue.

Should brands stop targeting AI Overviews because of the volatility?

Brands should not stop targeting AI Overviews because of the volatility, but they should change how they measure success. Single-snapshot checks ("did our URL appear in this AIO today") are unreliable. The right approach is to measure citation share across many repeat runs of priority prompts, prioritise the signals AI engines weight most heavily (earned media, structured content, FAQ schema, Bing visibility, entity strength), and track multiple AI surfaces in parallel because they rarely agree.

What is the strongest predictor of being cited consistently in AI search?

The strongest predictor of consistent AI citation share in 2026 is earned-media coverage in publications the AI engine already trusts. The University of Toronto's 2025 generative engine optimisation study described AI search's bias toward earned media over brand-owned content as "systematic and overwhelming." AuthorityTech research quantified the gap at approximately 325% more citations for equivalent earned-media coverage versus brand-owned content. Earned media is also volatility-resistant: it stays valuable regardless of how the AI engine reshuffles URLs day to day.

How quickly do AI citation patterns change?

AI citation patterns can change dramatically within weeks. Semrush analysis showed ChatGPT's Reddit citation share dropping from approximately 60% to 10% over roughly six weeks in late 2025, while Forbes, PRNewswire, and Medium absorbed the displaced share. Continuous tracking is essential because static optimisation strategies will not match the engines' evolving source preferences.

What does AI Mode's "query fan-out" technique mean for SEO?

AI Mode's query fan-out technique breaks each user query into multiple sub-queries that the system runs in parallel, then synthesises into a single response. The implication for SEO is that ranking for a single keyword is no longer sufficient. Pages need to be retrievable across the related cluster of sub-queries that AI Mode is likely to generate from a high-intent prompt, which favours topical depth and structured content over single-keyword targeting.

Is Bing more important than Google for ChatGPT visibility?

Yes, for ChatGPT specifically. ChatGPT pulls primarily from Bing's index when performing live web searches, with additional signals from OpenAI training data. Independent research has found that only around 12% of ChatGPT citations match URLs on Google's first page. A page ranking #15 in Google but #5 in Bing may be cited by ChatGPT more often than a page ranking #1 in Google but absent from Bing's top results. Bing Webmaster Tools is free and remains underused.

How can brands measure AI visibility consistently when results vary so much?

Brands measure AI visibility consistently by running the same prompts multiple times per day across each AI engine and reporting share of voice as a percentage across runs, not as a binary "appeared / did not appear." Purpose-built AI visibility platforms such as KIME automate this measurement across 10+ AI engines on a daily schedule, with citation source tracking on every plan tier.

What is the highest-leverage action a brand can take this quarter to improve AI visibility?

The highest-leverage action a brand can take this quarter is to identify the 5 to 10 publications each major AI engine cites most often in their category and pitch original data, expert commentary, or feature stories to those publications. Earned media outperforms brand-owned content by approximately 325% for AI citation rates, and that lift is volatility-resistant. The next-highest actions are restructuring the top 20 brand-owned pages with answer-first formatting and adding FAQ schema.

Related guides

Start a free trial of KIME → and see exactly how volatile your visibility is across each AI search engine.

This guide was written by the KIME team and synthesises publicly available research from SE Ranking, Semrush, Serpstat, Analyze AI, Zyppy, Authoritas, AuthorityTech, the University of Toronto, Search Engine Land, Search Engine Journal, and TechEdgeAI's 2026 AI Platform Citation Source Index. Citation patterns and platform behaviours change frequently; verify current data from each source before making strategic decisions.

Source list (linked inline above):

SE Ranking, "AI Mode research: Volatility, source patterns, and differences from AIO and organic results" (June 2025): https://seranking.com/blog/ai-mode-research/
SE Ranking, "70+ AI Search Stats for 2026" (January 2026): https://seranking.com/blog/ai-statistics/
Search Engine Journal, "Study: Google AI Mode Shows 91% URL Change Across Repeat Searches" (July 2025): https://www.searchenginejournal.com/study-google-ai-mode-returns-largely-different-results-across-sessions/550249/
Semrush, "Exploring URL Volatility in Google's AI Overviews" (Mordy Oberstein, November 2024): https://www.semrush.com/blog/url-volatility-ai-overviews/
Serpstat, "Inside Google's AI Overviews: Patterns, Volatility, and Ranking Signals" (January 2026)
TechEdgeAI, "AI Platform Citation Source Index 2026" (May 2026): https://techedgeai.com/ai-platform-citation-source-index-2026-shows-reddits-surge-and-a-new-era-of-volatile-ai-generated-answers/
Search Engine Land, "4 signals that now define visibility in AI search" (April 2026): https://searchengineland.com/visibility-ai-search-signals-475863
SEO First, "Google Volatile March 2026 Update Outlined": https://www.seofirst.com/index.php/google/google-volatile-march-2026-update-outlined/
Analyze AI, ChatGPT vs Claude vs Perplexity citation study (Nov 2025–Jan 2026): https://tryanalyze.ai/
AuthorityTech, Machine Relations earned-media research (2025–2026): https://authoritytech.io/
University of Toronto, generative engine optimisation study (arXiv, September 2025)