GEO Beyond ChatGPT: Do Claude, Perplexity, and Gemini Still...

Yes, but the opportunity is not “more traffic than ChatGPT.” The better question is whether non-ChatGPT platforms send more research-heavy referrals, cite different pages, or convert at a higher rate than their visit volume suggests. Goodie’s 2026 table shows that Claude and Perplexity have much smaller platform visit shares than ChatGPT, yet their normalized referral shares are much higher than those visit shares. That is the growth signal worth testing.

Short voiceover summary: why GEO teams should test Claude, Perplexity, Gemini, Copilot, and Grok separately instead of hiding them under “other AI.”

Key Takeaways

GEO teams should treat ChatGPT as the scale baseline and treat Claude, Perplexity, Gemini, Copilot, and Grok as separate opportunity rows. The public benchmark in the Goodie table does not prove every brand should invest equally in every platform. It does show that platform traffic and outbound referral behavior do not move in a straight line.

ChatGPT still owns the largest visible baseline. In the Goodie table, ChatGPT has 64.4% of measured platform visits from January to April 2026 and 60.8% of recent normalized referral share.
Claude is the clearest referral-efficiency outlier in the table: 1.29% platform visit share, but 18.0% normalized referral share.
Perplexity is smaller in total volume, with 1.85% platform visit share, but its 7.1% referral share suggests a stronger source-click or research pattern than raw visits imply.
Gemini is not the same kind of opportunity. It has large platform usage at 29.0% visit share, but only 10.3% referral share in the table. That makes it a scale-and-growth watchlist item, not automatically a high-referral-efficiency channel.
Copilot and Grok should remain in the dashboard, but with different labels. Copilot has a measurable 3.9% referral share in the table while platform visits were not measured. Grok had 3.5% visit share and about 0% referral share.
Measurement must separate three signals: platform visits, AI referral clicks, and AI-influenced demand. Google’s generative AI optimization guide and Search generative AI control documentation make the caveat important: some AI exposure lives inside Google Search reporting, not a clean referrer row.

The practical next step is a monthly non-ChatGPT GEO test. Pick 20 prompts per priority category, test them across ChatGPT, Claude, Perplexity, Gemini, and Copilot, record whether your brand appears, which URLs are cited, and whether the cited page receives any visible referral or follow-up demand. A free AI visibility snapshot can be the first baseline before you build a larger tracking sheet.

Why ChatGPT Alone Is Not Enough For GEO

ChatGPT is still the first platform most teams should measure, but it is not the whole GEO market. A ChatGPT-only report tells you where the largest visible AI source is sending traffic. It does not tell you whether smaller platforms are stronger for research, comparison, supplier discovery, technical validation, or late-stage purchase questions. The Goodie table makes the reason visible. ChatGPT’s visit share and referral share are close: 64.4% versus 60.8%. Claude and Perplexity are different. Their visit shares are tiny, but their referral shares are much larger. That mismatch is exactly what a GEO team wants to find, because a smaller source can be worth work if it sends users who are checking sources, comparing options, or validating a recommendation.

Platform group	What it is good for in a GEO report	What not to assume
ChatGPT	Scale baseline, brand recall, broad answer coverage, cited pages from high-volume prompts	That it captures every AI-influenced visit
Claude	Research and explanation-heavy prompts, B2B-style evaluation, high referral share relative to visits in the Goodie table	That direct volume will be large enough for every site
Perplexity	Source-led research, comparison journeys, pages users click to verify details	That growth is guaranteed in every region or category
Gemini	Growth watchlist, Google ecosystem behavior, possible spillover through Search	That visit share will become clean referral traffic
Copilot	Work-context discovery and professional workflows	That platform visits can be benchmarked cleanly from the table
Grok	Awareness watchlist where audience fit exists	That platform usage alone means outbound traffic

This is also why “GEO” should not become a platform-by-platform content rewrite. Google’s official guidance says generative AI features in Search depend on core Search systems and useful pages. OpenAI’s ChatGPT Search help page says search answers can show citations and a Sources panel. The shared lesson is simple: build pages that can be retrieved, understood, cited, and clicked. Then test which AI surfaces actually use them. For a content lead, the immediate action is to stop reporting “AI traffic” as one row. Split it into platform, cited URL, prompt group, landing page type, and outcome. If your current analytics setup cannot do that yet, start with prompt tests and visible referrers, then add conversion quality later.

What The Goodie Table Actually Shows

The table is not saying Claude or Perplexity are bigger than ChatGPT. It is saying platform usage and referral output are different behaviors. ChatGPT has the largest total use and the largest referral share. Claude and Perplexity have limited platform visits but disproportionately high normalized referral share, which makes them good candidates for deeper GEO testing.

Visit share vs referral share infographic

Platform	Total visits Jan-Apr 2026	Visit share	Recent normalized referral share	GEO interpretation
ChatGPT	16.59B	64.4%	60.8%	Scale baseline; keep measuring, but do not stop there
Gemini	7.47B	29.0%	10.3%	Large platform use; referral behavior is weaker than visit share
Grok	904M	3.5%	about 0%	Watch only when audience fit is clear
Perplexity	476M	1.85%	7.1%	Small reach, stronger source-click signal
Claude	333M	1.29%	18.0%	Strongest referral-share overperformance in the table
Copilot	Not measured	n/a	3.9%	Keep as a separate row; do not bury under “other”

The difference between visit share and referral share is the article’s main point. Visit share measures how much people use the AI platform. Referral share measures how often measurable outbound traffic comes from that platform. A platform can have many users who do not click sources, and a smaller platform can have users who click because they are doing research. Goodie’s public report is a B2B-oriented benchmark, so it should not be copied directly into an ecommerce, SaaS, or marketplace forecast. Use it as a measurement lesson. If your audience asks comparison questions, policy questions, technical questions, or supplier-evaluation questions, smaller research-heavy platforms may matter earlier than their raw visit share suggests. The safe conclusion is narrow but useful: do not rank AI platforms only by total usage. Rank them by usage, referral behavior, citation presence, and business outcome together.

Where Claude And Perplexity Create Opportunity

Claude and Perplexity create opportunity when the buyer needs evidence. These platforms are not a replacement for ChatGPT scale. They are more interesting for prompts where the user wants sources, comparisons, explanations, and confidence before taking the next step. That is why their referral share can be more important than their total platform share. In the Goodie table, Claude has 1.29% visit share and 18.0% referral share. Perplexity has 1.85% visit share and 7.1% referral share. Those numbers should trigger a test, not a budget shift by themselves.

Opportunity signal	Claude	Perplexity	What to test next
Referral share above visit share	Very strong in the Goodie table	Strong in the Goodie table	Track source clicks from comparison and research prompts
Research workflow fit	Strong for explanation, evaluation, and B2B questions	Strong for source-led answers and verification	Build pages that answer “why this option” and “how to compare”
Total volume	Still limited	Still limited	Use conversion quality, not raw sessions, to decide priority
Best content candidates	Method pages, comparison pages, evidence-backed guides, case-style explainers	Source-rich guides, product/category comparisons, data pages, FAQ hubs	Record cited URLs and update pages that are already close
Main caveat	High referral share may not repeat in every vertical	SE Ranking reports regional and time-based volatility in AI traffic share	Re-test monthly and compare by category

An anonymized ecommerce sample points in the same direction without proving a universal rule. In that sample, ChatGPT produced 82.0% of tracked AI referral UV and 55.0% of GMV. Perplexity produced 7.0% of UV and 35.6% of GMV, with a higher order rate. That does not mean Perplexity is always better. It means “small platform, high-intent visit” is a real reporting pattern teams should be able to detect. The action is to create a Claude/Perplexity prompt set before rewriting content. Use prompts like “best options for [category],” “compare [product type] for [use case],” “what should I check before buying [product/service],” and “which sources explain [technical requirement].” If your pages are cited but not clicked, improve the title, snippet-like answer block, comparison table, and supporting evidence. If you are not cited at all, the gap is visibility, not conversion.

How Gemini, Copilot, And Grok Fit Into The Watchlist

Gemini, Copilot, and Grok should stay in the dashboard, but they need different labels. Gemini is a growth-and-scale watchlist item. Copilot is a professional workflow watchlist item. Grok is an audience-fit watchlist item. None of them should be hidden inside “other AI,” because early changes are easy to miss once reporting is aggregated. SE Ranking’s 2026 AI traffic research says Gemini grew quickly from a small base, and its separate Gemini-versus-ChatGPT analysis reports that Gemini overtook Perplexity by its measurement in early 2026. That does not mean Gemini sends better visitors today. It means the platform is large enough and connected enough to Google’s ecosystem that GEO teams should watch it separately.

Platform	Dashboard label	Main question	Practical test
Gemini	Growth and Search-adjacent watchlist	Are Gemini and Google AI surfaces changing brand discovery?	Compare prompt visibility with Search Console changes and organic landing pages
Copilot	Work-context watchlist	Does the audience use Microsoft workflows for research?	Test procurement, B2B, documentation, and how-to prompts
Grok	Audience-fit watchlist	Does the category’s audience overlap with Grok’s active users?	Track only if prompts show mentions, citations, or competitor presence

Gemini is the easiest to overread. In the Goodie table, Gemini has 29.0% visit share but 10.3% referral share. That means large platform use did not translate into equal outbound referrals in that benchmark. The right conclusion is not “ignore Gemini.” It is “separate visibility from referral traffic.” A page can influence discovery inside a Google or Gemini experience without leaving a clean referrer trail. Copilot has the opposite measurement problem: the table lists referral share, but platform visits were not measured. Keep it in reporting when your buyers live in work tools, especially B2B teams, developers, operators, and procurement users. Grok should be watched only where audience fit is plausible or where prompt tests show repeated mentions. The clean reporting rule is to give each platform one of four labels: scale baseline, referral-efficiency opportunity, growth watchlist, or audience-fit watchlist. That keeps the team from treating all non-ChatGPT sources the same.

How To Research GEO Beyond ChatGPT

Researching GEO beyond ChatGPT is a monthly measurement habit, not a one-time article rewrite. The goal is to find where non-ChatGPT platforms mention the brand, cite the site, send visible referral traffic, or influence follow-up demand. The workflow should be small enough to repeat and structured enough to compare month over month. Use this workflow:

Pick 3-5 priority categories or buyer questions.
Write 10-20 prompts per category. Include discovery prompts, comparison prompts, risk prompts, and “what should I check before buying” prompts.
Run the prompts in ChatGPT, Claude, Perplexity, Gemini, and Copilot. Add Grok only where audience fit is real.
Record brand mention, rank/order of mention, cited URLs, competitor URLs, answer framing, and whether the answer includes source links.
Match visible referrers in analytics to the same time window. Use normalized platform names such as ChatGPT, Perplexity, Gemini, Claude, Copilot.
Add business quality metrics where available: leads, trial starts, orders, GMV, pipeline, revenue per visit, or qualified lead rate.
Re-test the same prompt set monthly, then add or remove prompts based on category changes.

Field	Why it matters
Platform	Prevents non-ChatGPT movement from being hidden
Prompt group	Separates discovery, comparison, and purchase-intent behavior
Brand mentioned	Measures answer visibility
URL cited	Shows which page is trusted enough to support the answer
Competitor cited	Reveals citation gaps and page types to study
Visible referral	Connects AI answer behavior to analytics
Business outcome	Prevents teams from optimizing only for visits
Attribution note	Keeps direct, organic, and AI-influenced demand separate

Do not make the first version too complex. A spreadsheet with 50 prompts and five platform columns is enough to start. Once the pattern is clear, move the repeated checks into a dashboard. A page audit can then help decide whether the cited or missing page has crawlability, structure, internal-link, or answer-block problems.

What Content Gets Recommended By Research-Heavy Platforms

Research-heavy platforms are more likely to reward pages that help users verify a choice. That usually means comparison tables, evidence-backed guides, category explainers, data pages, policy pages, and detailed product or solution pages. Thin brand pages can still appear for navigational questions, but they rarely satisfy comparison or validation prompts. The content goal is not to write for one AI bot. It is to create pages that a retrieval system can understand and a cautious user can trust. Google’s official AI optimization guidance points back to core Search quality, crawlability, helpful content, and clear media. The same discipline helps answer engines because they need concise claims, sourceable facts, and page structure.

Prompt intent	Page type that can win	What the page needs
“What is the best option for X?”	Comparison guide	Selection criteria, table, caveats, who each option fits
“How do I choose X?”	Buyer guide or category guide	Step-by-step decision rules, definitions, examples
“Is X safe / reliable / compliant?”	Evidence or policy page	Source-backed statements, dates, certifications, limits
“Compare A vs B for Y”	Use-case comparison page	Clear tradeoffs, not just feature lists
“What should I check before buying X?”	Checklist page	Practical checks, warning signs, next actions
“Where can I find specs or requirements?”	Product/detail page	Structured specs, FAQs, images, downloadable evidence if useful

For Claude and Perplexity testing, start with pages that already have some evidence depth. Add short answer blocks near the top, then support them with tables, examples, and source links. If the page is a product page, add a plain-language summary of who it fits, who it does not fit, and which questions it answers. If the page is a guide, add comparison criteria and a dated source statement. One important caveat: do not flood the site with near-duplicate “AI answer pages.” A better pattern is to improve the pages that already match buyer intent and connect them with internal links. Convertos’ GEO tutorials can be used as the next-step hub for teams building this into a repeatable content workflow.

Common Measurement Mistakes And Attribution Limits

The most common mistake is to treat non-ChatGPT GEO as a traffic-volume race. That misses the point. The opportunity often shows up as disproportionate referral share, higher-intent sessions, new cited URLs, or assisted demand that later appears as direct, organic, or branded search. Measurement must be cautious enough to avoid false precision.

Mistake	Why it causes bad decisions	Better rule
Ranking platforms only by total visits	Smaller platforms with stronger research behavior disappear	Compare visit share, referral share, citation rate, and outcome
Treating referral share as revenue share	Referral behavior does not guarantee conversion	Add order rate, qualified lead rate, GMV, or pipeline quality
Hiding all non-ChatGPT platforms under “other”	Early signals are averaged away	Keep separate rows for Claude, Perplexity, Gemini, Copilot, and Grok
Assuming Google AI sends a clean referrer	Google AI features live inside Search systems	Pair analytics with Search Console and prompt tests
Rewriting pages before prompt testing	The team may fix pages AI systems never consider	Test prompts, record cited URLs, then optimize
Treating one benchmark as universal	Platform behavior varies by category, region, and user task	Use public data as a hypothesis, then test your own site

The attribution caveat matters most. A person can see a brand in an AI answer, copy a URL, search the brand later, or click through Google after reading an AI-generated summary. That path may never show as a clean AI referral. Google’s documentation around AI Overviews and AI Mode reinforces this point because those experiences sit within Search reporting rather than a separate assistant referrer. The fix is to label every metric by confidence. “Visible referral” is high confidence. “Cited in AI answer” is medium confidence. “AI-influenced demand” is directional unless you have a controlled test. That language keeps the report useful without overstating what the data can prove.

FAQ

These questions come from the prompt cluster behind this article and the recurring search intent around AI referral measurement, Claude/Perplexity opportunity, Gemini growth, and non-ChatGPT GEO prioritization.

Which AI platforms beyond ChatGPT are most worth GEO work?

Start with Claude, Perplexity, and Gemini, then add Copilot if your audience works in Microsoft-heavy workflows. Source signal: the Goodie table highlights Claude/Perplexity referral-share overperformance, while SE Ranking reports Gemini growth from a small base.

Why do Claude and Perplexity have higher referral share than visit share?

The likely reason is user task. Source signal: the Goodie table shows the visit-share versus referral-share gap, but your own prompt tests and analytics should confirm whether users are researching, comparing, or verifying sources.

Is Perplexity worth effort if traffic is limited?

Yes, if its visitors or citations show stronger intent. Source signal: the anonymized sample used for this article showed a smaller Perplexity row producing a much higher GMV share than its UV share; that is directional evidence, not an industry average.

How should teams judge Gemini’s growth opportunity?

Measure Gemini separately from generic Google traffic where possible, but do not expect visit share and referral share to match. Source signal: SE Ranking’s Gemini analysis and Google’s Search documentation both point to growth and attribution caveats.

How do you measure non-ChatGPT AI traffic?

Normalize visible referrers, record platform-specific prompt tests, track cited URLs, and add conversion quality. Source signal: related questions around AI traffic measurement ask how to separate visible referrers from AI-influenced demand.

Which pages are more likely to be cited by research-heavy platforms?

Pages that help a user verify a decision: comparison guides, buyer checklists, category explainers, data pages, policy pages, and detailed product pages. Source signal: the prompt cluster behind this article focuses on comparison, risk, and buying-check questions.

Source Statement

This article uses public AI search traffic research, official platform documentation, and one anonymized analytics sample. The public benchmark at the center of the article is Goodie’s 2026 AI search traffic table, which compares platform visits from January to April 2026 with recent normalized referral share. The article also references SE Ranking’s AI traffic studies, Google’s generative AI Search documentation, Google’s Search generative AI control page, and OpenAI’s ChatGPT Search help page. The anonymized sample is used only to explain a reporting pattern: a smaller AI platform row can sometimes produce stronger business quality than its visit share suggests. It is not presented as a market average. It does not name the source brand, source domain, internal team, original file, customer base, or any private identifier. AI search behavior changes quickly. Platform UX, citation behavior, source panels, Google AI reporting, and browser/app referrers can all change. Teams should re-check source documentation and their own analytics before making roadmap or budget decisions.

GEO Beyond ChatGPT: Do Claude, Perplexity, and Gemini Still Offer Growth Opportunities?