Why does my brand show up in ChatGPT but not Perplexity?

Because they read from different indexes and apply different source-selection rules. ChatGPT (reportedly Bing-based) reaches broadly across the web and leans on Wikipedia and Reddit, while Perplexity blends its own index with Google's and reranks hard for authority, recency, and clean structure. A page that fits one engine's answer style can be a poor match for the other, so presence in one is no guarantee of the other.

Should I track AI visibility per platform or as one blended score?

Per platform, always. A single averaged score hides the surface that's failing: being cited in eight of ten ChatGPT checks and none of your Perplexity checks blends to a comfortable-looking midpoint while you're invisible to an entire audience. Keep each assistant in its own column so you can see exactly which one is ignoring you and act on that specific gap.

Which AI assistants should I measure my brand in?

Start with the ones your buyers use: ChatGPT and Google's AI Overviews reach the widest audiences, Perplexity skews toward researchers and technical buyers, Copilot toward Microsoft-ecosystem enterprises, and Claude toward developers and analysts. Because each uses a different backend, cover all five if you can — a gap on any one is a gap you can't see until you measure it directly.

Can I optimize once and rank everywhere across assistants?

No. Universal fundamentals help everywhere — cited sources, statistics, clear structure, schema, and crawler access. But each surface also has its own lever: Bing indexing for Copilot, Brave Search visibility for Claude, FAQ schema and publishing cadence for Perplexity, third-party footprint for ChatGPT. Fix the fundamentals first, then tune the surface-specific gaps your measurement reveals.

Why your AI visibility differs across AI assistants

Guide · AI Visibility · 5 min read · last verified 2026-07-25

Reviewed before publication Editorial board Independent commercial review

In shortYour brand can be cited by one AI assistant and invisible in another because each uses a different index, source-selection logic, and recency weighting. Measure per platform, never average into one score.

The short answer: different assistants use different indexes and different rules

The same brand can be cited confidently by ChatGPT and be absent in Perplexity — or dominate Google's AI Overviews while never surfacing in Claude. It's not a bug in your content. Each assistant reads from a different search index, applies its own source-selection logic, weights recency differently, and leans on different training data. The takeaway: treat every assistant as its own channel, measure each separately, and never average them into a single "AI visibility score."

Four reasons the same brand appears in one assistant and not another

Cross-assistant variance comes down to four moving parts, all observed/reported — no platform publishes its ranking formula, so treat each mechanic as directional, not certain.

Different indexes and backends. An assistant can only cite what its backend has indexed, and the backends differ — Bing, Google, Brave, or a blend (see the table below). If Bing hasn't indexed a page, Copilot can't cite it, however well it ranks on Google.
Different source-selection tilts. With overlapping indexes, engines still favor different sources — some reach broadly across the web, others prefer a short list of high-authority, cleanly structured pages.
Different recency weighting. Some assistants prefer freshly updated pages; others lean on established sources. One analysis reports ChatGPT cites recently updated content much more often than older pages — a refresh can flip your presence on one engine but not another.
Different grounding. Answers blend live retrieval with training memory; when memory wins, older "canonical" sources (an established encyclopedia entry, say) can crowd out a newer, better page.

A per-assistant map of where each one looks

Read it as a diagnostic starting point: every row is reported/observed behavior that shifts as platforms update.

Assistant	Reported search backend	Observed source-selection tilt	What that means for your brand
ChatGPT (search)	Bing-based web index	Wide range of pages, not just top-ranked; heavy on Wikipedia and Reddit; favors recent updates	A strong third-party footprint (Wikipedia, Reddit) can matter as much as your own site
Perplexity	Own index blended with Google's, multi-pass reranking	Authority, recency, clean structure; rewards FAQ schema and self-contained paragraphs	Well-structured, frequently published, source-cited pages win here
Google Gemini / AI Overviews	Google's index plus the Knowledge Graph	Leans on existing Google / E-E-A-T signals and structured data	Your Google footprint and schema markup largely carry over
Microsoft Copilot	Bing's index	Bing-indexed, authoritative sources; LinkedIn and GitHub presence reportedly help	No Bing index, no citation — submit to Bing Webmaster Tools
Claude (web search)	Brave Search	Very selective; rewards factual density and clear attribution	Check Brave Search visibility; precise, dated, sourced content wins

The fragmentation is real, not random — which is why understanding how AI assistants choose their sources is the foundation for reading these differences.

Why a single averaged score hides the gap that matters

Rolling five assistants into one blended number is the most common measurement mistake — dangerous because it looks reassuring. Cited in eight of ten ChatGPT checks and none of your Perplexity checks? Blend them and you get a middling-but-fine "five of ten" — while you're invisible to every buyer who researches in Perplexity. Averaging trades away the one thing you can act on: which surface is failing, and why.

This is why AI answers are fragmented, not winner-take-all: no single source wins everywhere, and leaning on one channel is exactly what makes visibility fragile. The metrics that matter are per-platform citation presence and share of voice against named competitors, surface by surface.

What to do: a per-assistant action list

Lead with measurement; then fix the lever the failing surface rewards.

Measure each assistant separately. Run the same question set through ChatGPT, Perplexity, Gemini/AI Overviews, Copilot, and Claude, recording per platform whether you're cited, who else is, and which page they used. Keep the surfaces in separate columns — never collapse them.
Diagnose the gap per platform. Absent from Copilot? Check Bing indexing. Perplexity? Check structure, FAQ schema, and publishing cadence. ChatGPT? Check third-party footprint and how recently the page was updated. Claude? Check Brave Search visibility and factual density.
Fix the lever that surface rewards. According to the Princeton GEO study (KDD 2024, tested on Perplexity), a handful of edits reliably move answer-engine visibility:

Move	Reported visibility effect	Applies best to
Cite authoritative sources	+40%	All; strongest on Perplexity, Claude
Add specific statistics	+37%	Research-style answers
Add expert quotations	+30%	Authority-weighted surfaces
Authoritative (non-salesy) tone	+25%	All surfaces
Improve clarity / fluency	+15–30%	All surfaces
Keyword stuffing	−10% (actively hurts)	Avoid everywhere

Treat generalization beyond Perplexity as a hypothesis — the effect sizes were measured on one engine — but "cite sources, add statistics, write clearly, drop the keyword stuffing" is a safe bet across all of them.

Address the surface-specific fundamentals. Allow every AI crawler in robots.txt (GPTBot, PerplexityBot, ClaudeBot, Google-Extended, Bingbot); submit to Google Search Console and Bing Webmaster Tools; add FAQPage and Article schema; and build third-party presence — industry analyses of AI citations, according to those studies, put Wikipedia at roughly 7.8% of ChatGPT citations and Reddit around 1.8%, and that brands are cited more via third-party sources than their own domains.
Re-measure on a locked benchmark. Change the question set between checks and you can't tell whether a movement was your work or your measurement — freeze the questions, competitors, and cadence so the deltas mean something.

Honesty about what you cannot see

AI source selection isn't fully observable. Platforms don't publish their ranking formulas and change them without notice, so a mechanic that holds this quarter may shift next, and no one can guarantee placement in a named assistant. What you can do is measure real answers, label each finding by confidence (measured from your own checks, derived from patterns, hypothesis where you're inferring), and re-run on a fixed benchmark so trend lines stay honest. According to industry estimates, AI Overviews already appear in roughly 45% of Google searches — so any surface you aren't measuring is one quietly shaping buyer perception.

Where Magrios fits

This per-assistant reality is the core of how Magrios works. Instead of one blurred score, it measures your brand's citation presence and share of voice on each surface — ChatGPT, Perplexity, Gemini, Copilot, Claude — against the competitors you name, then re-measures on a locked benchmark so every movement is real. Each finding links back to the actual AI answer it came from, so "absent in Perplexity" is a claim you can click into and verify — treating each citation surface as its own measurable channel. Measure the gaps per platform, act on the surface that's failing, then re-measure — that loop, not a single headline number, is what turns cross-assistant variance from a mystery into a to-do list.