Is Optimising for ChatGPT Enough? Why a Single-Assistant GEO Strategy No Longer Works

Three shifts happened in roughly twelve months.

The market went from one dominant player to a competitive field. ChatGPT's worldwide AI web-traffic share dropped from about 76% to roughly 53%. Gemini moved to around a quarter of the category. Claude grew fastest proportionally, overtaking Perplexity in referral sessions in March 2026 after growing substantially. Perplexity's referral sessions fell sharply from their earlier peak.

Referral behaviour differs sharply by platform. In measurable standalone AI referrals, ChatGPT still dominates by volume — one analysis of 6.77 million AI-driven sessions across 166 sites found ChatGPT sending 92.4% of trackable standalone AI referral traffic. But share of referrals and share of usage are different metrics, and optimising only where the clicks are visible misses where the recommendations happen.

Google's AI surfaces outweigh all standalone assistants combined. Google's AI Overviews and AI Mode together produce more AI-influenced traffic than ChatGPT, Claude, Gemini, Perplexity and Copilot combined. A GEO strategy that treats "AI search" as synonymous with ChatGPT ignores the largest surface entirely.

The strategic conclusion holds regardless of which vendor's decimal you trust: a ChatGPT-only approach now leaves a large slice of AI-referred and AI-influenced visitors unaddressed.

This is the finding with the most operational consequence, and the least coverage.

Analysis of AI citation behaviour found that only about 11% of citations overlap between major AI platforms, and that citation sets change roughly 50% month to month.

Two implications follow directly.

Being cited by one platform tells you little about the others. If you appear in ChatGPT answers for your category, that is genuinely valuable — and it is weak evidence about whether Gemini, Perplexity or Claude name you for the same query. Each platform draws on different sources with different weightings. You have to check each one separately.

A single measurement is a snapshot, not a trend. With citation sets shifting by half each month, one check tells you almost nothing about your position. This is why monthly tracking matters more than depth of any single audit — consistency of measurement beats precision in one reading.

There is a related finding worth flagging honestly, because it complicates a common assumption: independent analyses report a sharp decoupling between traditional top-10 rankings and AI citations, with a large share of AI citations now coming from URLs outside the organic top results. Ranking well is still helpful and still the most reliable foundation — but it is no longer sufficient, and it no longer predicts citation the way it once did.

Beyond citing different sources, the assistants have different presentational preferences, which affects how you structure content.

Broadly: ChatGPT tends toward comparison tables, Copilot toward narrative summaries, and Perplexity toward extracting from bullet lists. Covering multiple formats on the same page — a table, a short narrative summary, and a scannable list — maximises the chance a given page is usable across platforms rather than optimised for one.

Citation rates also vary sharply by category. Overall citation rates remain modest — one analysis puts the general rate below 7%, meaning most content still does not make it into AI answers verbatim. But categories built on comparable, checkable facts — travel, automotive, and by extension anything with specifications, prices, or clear comparisons — see citation rates three to five times the average.

For a Singapore business, that is an actionable signal: content built around concrete, verifiable facts (pricing bands, specifications, local requirements, comparison criteria) is structurally more citable than general commentary, across every platform.

You do not need a separate strategy per assistant. You need one strategy that does not assume a single platform, plus per-platform verification.

Step 1: Establish which platforms matter for your buyers. For a Singapore B2B firm, that usually means ChatGPT, Gemini (given Google's integration into everyday search) and Claude, with Perplexity relevant for research-heavy buyers. Consumer-facing businesses should weight Google's AI surfaces heavily.

Step 2: Check each platform separately. Take your ten most commercially important queries and run them across each relevant assistant. Log for each: are you named, is the description accurate, and which competitors appear instead. Given the 11% overlap, results will differ meaningfully — that variance is the point of the exercise.

Step 3: Cover multiple content formats. On your key pages, include a comparison table, a narrative summary paragraph, and a scannable list. This is cheap to do and materially widens which platforms can use the page.

Step 4: Lead with checkable facts. Given that fact-dense, comparable content earns citation rates several times the average, prioritise content containing specific figures, defined criteria and verifiable claims over general thought leadership.

Step 5: Track monthly, not quarterly. With citation sets changing about 50% month to month, quarterly checks will mislead you. A twenty-minute monthly log across platforms is more useful than an annual deep audit.

Step 6: Do not neglect the technical layer. Different assistants use different crawlers — GPTBot and OAI-SearchBot for OpenAI, ClaudeBot for Anthropic, PerplexityBot for Perplexity, Google-Extended for Google's AI features. Blocking any one of them removes you from that platform specifically, and this is easy to do accidentally through CDN or robots.txt configuration.

A practical note, because multi-platform sounds like multiplied cost. It usually is not.

The content work is largely shared: fact-dense, well-structured, multi-format pages serve every platform. The technical work is shared: crawler access, schema, clean structure. The genuinely per-platform work is verification and monitoring, which is measurement time rather than production spend.

What does change is the assumption behind your reporting. An agency reporting only ChatGPT visibility is reporting on roughly half the picture. When commissioning GEO work in Singapore, ask specifically which platforms are tracked, how often, and whether the citation log covers all of them — because with 11% overlap, "we track AI visibility" without specifying platforms is not a meaningful commitment.

Q1: Is ChatGPT still the most important AI platform for visibility?

It remains the largest single assistant by both web and app usage, and it dominates measurable standalone referral traffic — one analysis of 6.77 million AI-driven sessions found ChatGPT sending 92.4% of trackable standalone AI referrals. But its share of overall AI web traffic fell from roughly 76% in June 2025 to about 53% by May 2026, not because ChatGPT shrank but because the category grew faster around it. Gemini now holds roughly a quarter, and Claude has grown fastest proportionally. Most importantly, Google's AI Overviews and AI Mode together produce more AI-influenced traffic than all standalone assistants combined — so ChatGPT alone is no longer a sufficient scope.

Q2: If I appear in ChatGPT answers, will I appear in Gemini and Perplexity too?

Not reliably. Analysis of citation behaviour found only around 11% of citations overlap between major AI platforms, meaning each draws on different sources with different weightings. Being cited by one platform is genuinely valuable but is weak evidence about the others — you have to check each separately. Compounding this, citation sets change roughly 50% month to month, so even a positive check on one platform is a snapshot rather than a stable position. The practical response is to run your key queries across each relevant assistant monthly and log the results per platform.

Q3: Does ranking well on Google still guarantee AI citations?

No, and this has changed. Independent analyses report a sharp decoupling between traditional top-10 rankings and AI citations, with a large share of AI citations now coming from URLs outside the organic top results. Ranking well remains helpful and is still the most reliable foundation to build on — but it is no longer sufficient, and it no longer predicts citation the way it once did. This is why AI visibility now requires its own measurement rather than being inferred from ranking data, and why structural and off-site factors matter alongside conventional SEO.

Q4: How should I structure content to work across multiple AI platforms?

Cover multiple formats on the same page, because platforms have different presentational preferences: ChatGPT tends toward comparison tables, Copilot toward narrative summaries, and Perplexity toward extracting from bullet lists. Including a table, a short narrative summary, and a scannable list on your key pages is inexpensive and materially widens which platforms can use the content. Beyond format, prioritise fact density: overall citation rates remain below 7%, but categories built on comparable, checkable facts see citation rates three to five times the average — so content with specific figures, defined criteria and verifiable claims is structurally more citable than general commentary.

Q5: Does optimising for multiple AI platforms cost significantly more?

Usually not, because most of the work is shared. Fact-dense, well-structured, multi-format content serves every platform, and the technical foundations — crawler access, schema, clean page structure — are common across them. The genuinely per-platform work is verification and monitoring, which is measurement time rather than production cost. What should change is your reporting assumption: an agency reporting only ChatGPT visibility is covering roughly half the picture. When commissioning GEO work, ask specifically which platforms are tracked and how often, since with only 11% citation overlap, "we track AI visibility" without naming platforms is not a meaningful commitment.

Mayson AI tracks AI visibility across ChatGPT, Gemini, Claude, Perplexity and Google's AI surfaces for Singapore businesses — with per-platform citation logs rather than a single-assistant snapshot. If you want to know where you stand across all of them, book a consultation.

Figures reflect published research from Similarweb, Previsible and related analyses as at August 2026. Panel-based estimates vary by vendor and methodology.