Jonathon's AI Wiki

AI Answer Engines Compared — ChatGPT, Google AI Overviews, AI Mode, Perplexity and Gemini

10 min read

Source: Usage and traffic: ai-research/momentic-ai-chatbot-market-share-2026-06-30.md, ai-research/ppcland-genai-traffic-share-june-2026-2026-06-30.md, ai-research/searchenginejournal-gemini-gains-share-2026-06-30.md, ai-research/similarweb-gen-ai-market-winners-2026-06-30.md, ai-research/similarweb-gen-ai-stats-2026-06-30.md, ai-research/similarweb-most-used-ai-aug-2025-2026-06-30.md, ai-research/similarweb-2026-genai-brand-visibility-2026-05-19.md, raw/similarweb-2026-generative-ai-landscape-evolution-ai-search.pdf, ai-research/quickseo-google-ai-overviews-statistics-2026-2026-07-25.md (a vendor compilation of third-party figures). Citation studies: ai-research/sistrix-ai-citation-drift-2026-05-19.md, ai-research/ahrefs-ai-mode-vs-ai-overviews-730k-2026-05-19.md, ai-research/ahrefs-38pct-aio-top10-update-2026-05-19.md, ai-research/ahrefs-schema-ai-citations-study-2026-05-11.md, ai-research/seranking-google-self-citation-ai-mode-2026-05-19.md, ai-research/seranking-50-ai-mode-factors-2026-05-19.md, ai-research/similarweb-most-cited-domains-llms-2026-04-14.md, ai-research/digital-applied-1000-aio-citation-pattern-study-2026-04-26.md, ai-research/airops-fan-out-effect-citation-study-2026-04-13.md, ai-research/geo-16-framework-arxiv-kumar-palkhouski-2025-09.md, ai-research/zyppy-ai-citation-ranking-factors-2026-05-07.md. Experiments, tools and practitioner guidance: ai-research/ramp-marketing-to-ai-agents-2026-05-28.md, ai-research/similarweb-ai-chatbot-traffic-trackers-2026-06-30.md, raw/geo-playbook-devesh-paliwal.pdf.

Google AI Overviews, Google AI Mode, ChatGPT, Perplexity and the Gemini app are separate citation systems; where overlap has been measured, they cite mostly different sources for the same prompt. This page regroups the seo-content study cluster by engine, with each study’s scope beside its numbers, to answer where a dental-marketing agency should spend GEO effort and how to check it is working. Evidence is deep for the Google surfaces and ChatGPT and thin for Perplexity and the Gemini app.

Key Takeaways

  • Google has the reach; ChatGPT and Gemini have the chatbot traffic. Google reports 2B+ monthly AI Overview users (Alphabet disclosures, via QuickSEO). Of worldwide chatbot web visits in April 2026, ChatGPT had 54.7%, Gemini 27.4% and Perplexity 1.5% (Similarweb data via Momentic; web only, no API, app or embedded use).
  • Overlap is low. AI Overviews and AI Mode shared 13.7% of cited URLs (Ahrefs, 540K US query pairs, September 2025). SISTRIX measured domain Jaccard overlap of 0.17 (AI Overviews vs AI Mode) and 0.125 (AI Mode vs ChatGPT) across 82,619 prompts in six countries, December 2025 to April 2026.
  • Churn differs by engine (SISTRIX). AI Overview sources were unchanged for 17 weeks on 53% of prompts; AI Mode replaced 56% of domains weekly around a stable core; ChatGPT Search replaced 74% and usually had no core.
  • Authority signals disagree by engine. ^[ambiguous] DA correlates +0.61 with AI Overview citation (Digital Applied, 1,000 US AI Overviews, April 2026); AirOps found no positive DA correlation for ChatGPT (16,851 queries); SE Ranking’s model weights referring domains highest for ChatGPT (SHAP 1.21 vs 0.56 for AI Mode, December 2025).
  • Adding schema did not move already-cited pages on any engine. Ahrefs’ matched difference-in-differences test (1,885 already-cited pages adding JSON-LD, August 2025 to March 2026, vs 4,000 controls): ChatGPT +2.2% and AI Mode +2.4% (not significant), AI Overviews −4.6% (significant, small).
  • Perplexity and the Gemini app are barely studied. No source in this vault measures citation stability for either or isolates Gemini-app citations.

Comparison at a glance

EngineUsage (scope)What it tends to citeCitation stabilityEvidence depth
AI Overviews2B+ monthly users (Google, via QuickSEO); about 16% of US searches in Jan 2025, about 43% in May 2026 (Similarweb chart)YouTube is the top domain (Ahrefs, Mar 2026) or Reddit at 21% (QuickSEO) ^[ambiguous]; 37.9% of cited URLs rank top 10 (Ahrefs, 863K SERPs)Most stable: 53% of prompts unchanged over 17 weeks (SISTRIX)Deep, mostly US
AI ModeUS visits about 125M (Jun 2025), 320M (Mar 2026), 275M (May 2026) (Similarweb chart)google.com 17.42% (SE Ranking, Feb 2026) or fandom.com first at 7.16% (Similarweb, US, Jan–Feb 2026) ^[ambiguous]56% weekly churn; 1–5 domain core on 86.5% of prompts (SISTRIX)Deep; top-domain findings conflict
ChatGPT5.51B web visits, 54.7% share, Apr 2026 (Momentic)Wikipedia 13.15%, Reddit 11.97% (Similarweb, US, Jan–Feb 2026); 3–4 domains per answer (SISTRIX)Least stable: 74% weekly domain churn (SISTRIX)Deep
Perplexity155M web visits, 1.5% share, Apr 2026 (Momentic)No engine-level source data; GEO-16 pools its 492 of 1,702 citations with Brave and AI Overviews (B2B SaaS, Sep 2025)Not measuredThin
Gemini app2.76B web visits, 27.4% share, Apr 2026 (Momentic); 750M MAU, Q4 2025 (Alphabet, via QuickSEO)No Gemini-app citation study; per-URL retrieval cap favours content near the top (Petrovic, via Zyppy)Not measuredThin

The Similarweb trend values are read from unlabelled charts (US, desktop and mobile web). The evidence-depth column is this article’s judgement.

Google AI Overviews

  • Usage. Prevalence estimates run from about 21% to 60–65% of queries, depending on keyword set, market and device (QuickSEO). BrightEdge put the healthcare trigger rate at about 30% in February 2025 and 75%+ in February 2026 (via QuickSEO).
  • Ranking matters less than it did. Ahrefs (2 March 2026; 863K SERPs, 4M cited URLs) found 37.9% of cited URLs in the top 10 SERP blocks, against about 76% in July 2025. Ahrefs tentatively links the drop to Gemini 3, which has powered AI Overviews since January 2026, and to query fan-out.
  • Format. AI Overviews cite videos and core pages (homepages, category pages) nearly twice as often as AI Mode, and 11% carry no citation (Ahrefs, US, September 2025).
  • Source counts. ^[ambiguous] 4.2 citations per AI Overview, with the top 1% of domains taking 47% (Digital Applied: 1,000 AI Overviews, US-English desktop, April 2026); 11 domains (SISTRIX, six countries); 13.34 sources (QuickSEO). The units and markets differ.
  • Stability. SISTRIX: 8 of 11 domains persist; 53% of prompts unchanged, 28% with isolated changes, 19% churning 46% a week. “Once you’re in, you stay in.” Ahrefs’ earlier research, cited in its December 2025 study, found 45% of citations change between generations.
  • Value of a citation. ^[ambiguous] Both figures are attributed to Seer Interactive: 120% more organic and 41% more paid clicks for cited brands (quoted by Zyppy), or 35% more organic and 91% more paid clicks (quoted by QuickSEO).

Google AI Mode

  • Usage. ^[ambiguous] QuickSEO gives 100M+ monthly users in the US and India (Q2 2025) and 75M in total (December 2025), two figures that conflict.
  • Google cites itself, but by how much is disputed. SE Ranking (68,313 keywords, 20 niches, 1,321,398 citations collected 12 February 2026; market not stated): google.com was 17.42% of citations, up from 5.7% in June 2025. Of those links, 59% open Google results pages and 36.1% point to Google Business Profiles. Similarweb (about 600,000 citation events across ChatGPT and AI Mode, US, January–February 2026) ranks google.com fifth at 2.85%.
  • Reference and health sources over-index. Wikipedia is in 28.9% of AI Mode citations vs 18.1% for AI Overviews; health sites appear nearly twice as often as in AI Overviews (Ahrefs, US, September 2025). nih.gov and clevelandclinic.org are in Similarweb’s AI Mode top 20.
  • Predictors. Domain traffic from Google organic ranks first and referring domains second (SE Ranking, 50+ factors, December 2025).
  • Stability. SISTRIX: 14–16 domains per answer and 54–59% weekly churn in all six countries. A 1–5 domain core holds on 86.5% of prompts while 89% of the other domains rotate weekly. Core rates: YouTube 24%, news sites 1.4%.

ChatGPT

  • Usage. Similarweb’s tracker share fell from 86% (January 2025) to 64% (January 2026, via SEJ) and 52.7% (26 May 2026, via PPC Land).
  • Most answers carry no link. 6.8% of ChatGPT answers had web citations in May 2026, more than five times the June 2025 rate; health was 6.3% and travel 22.6% (Similarweb, US desktop).
  • What it cites. Wikipedia and Reddit together took about 25% of citations (Similarweb, US, January–February 2026). SISTRIX found documentation and institutional sources in the stable core, and 68% of core sources in English even for German prompts.
  • Retrieval rank dominates. Pages at retrieval rank 1 were cited 58.4% of the time vs 14.2% at rank 10 (AirOps: 16,851 queries run three times each through the ChatGPT interface, published April 2026).
  • Stability. SISTRIX: 74% of domains new each week (85% at URL level), with churn of 74% in Germany, 60% in the UK and 42% in France; the data covers only prompts with consistent source attribution. AirOps: 2.3% of page-query pairs were cited in all three runs.
  • Recent change. After the 7 May 2026 update added clickable brand links to answers, ChatGPT referrals rose 157.7% week over week, and the homepage share of referrals went from about 26–32% to about 60% (Similarweb).
  • Crawling is not citing. In Ramp’s five-week test on about 50 ramp.com pages, OpenAI SearchBot read the agent-addressed offer, but ChatGPT surfaced it zero times in 32 days.

Perplexity (thin evidence)

  • Usage. US share was 1.9% (Momentic, April 2026). Similarweb’s tracker had 2.0% in January 2026 and 1.3% in May 2026. PPC Land says API and mobile use partly explain the low web share, and reports about 780M queries a month in mid-2025.
  • Citation evidence:
    • GEO-16 (arXiv, September 2025; 70 prompts, 16 B2B SaaS verticals, English pages, one point in time) reports its correlations pooled across engines, with no Perplexity-only breakdown.
    • In Ramp’s test, Perplexity picked up the content by day 2 but kept it vague (“some channels offer bonuses”) until it named the offer on day 33.
    • Zyppy: citations drawn from training data, sometimes to URLs that no longer exist, are “more typical of ChatGPT and Perplexity.”
    • GEO Playbook (Paliwal; a practitioner deck with no sources): Perplexity crawls Reddit “in near real-time” and “indexes comments too.”

Gemini app (thin evidence)

  • Separate from AI Overviews. All three Google surfaces use Gemini, but QuickSEO advises tracking them separately.
  • Usage. Similarweb’s tracker share rose from 5% (January 2025, via SEJ) to 27.3% (May 2026, via PPC Land), and 2025 web visits grew 548%. App daily stickiness was 5.59%, against 43.87% for ChatGPT (Similarweb, August 2025). Gemini’s short prompts and few prompts per session are, in Similarweb’s words, “optimized for speed and execution” (US, June 2025 to May 2026).
  • Citations. Nothing Gemini-specific beyond Dan Petrovic’s per-URL retrieval-cap finding (via Zyppy). Gemini’s bot made few visits in Ramp’s logs.

How to monitor each engine

  • Log presence rates, not single snapshots. Track domain presence rather than URLs (SISTRIX), and run each prompt several times, as AirOps did. Run ChatGPT and AI Mode checks weekly and stable AI Overview checks monthly.
  • AI Overviews: SERP-feature trackers in Ahrefs, Semrush or SE Ranking. Search Console does not show whether you are cited inside an AI Overview (QuickSEO).
  • AI Mode: SE Ranking’s AI Mode tracker, Profound, Bluefish or Datos (QuickSEO).
  • ChatGPT, Perplexity, Gemini: a fixed prospect-question panel, plus server logs (Cloudflare files ChatGPT, Perplexity and Claude bots under “AI Assistants”, not “AI Search”, per Ramp). Ramp runs a weekly query set across Claude, ChatGPT, Perplexity and Gemini and compares the answers; the GEO Playbook suggests 20 queries a week plus the Perplexity API. For Gemini, QuickSEO lists Profound, BrandRank.AI, Otterly and Superlines.
  • Referral traffic: Similarweb’s account-gated trackers cover ChatGPT, Gemini, Perplexity, Claude, Grok and DeepSeek (Similarweb AI Search Intelligence).
  • KPI: report brand mentions as well as clicks. US AI-platform visits grew 28.6% from January 2025 to January 2026 while AI referrals stayed flat (Similarweb).

Try It

  1. Rank the engines for a dental client: AI Overviews and AI Mode first, because of their reach and the 75%+ healthcare trigger rate; ChatGPT second; Perplexity and Gemini monitor-only.
  2. Audit the Google Business Profile. 36.1% of google.com links inside AI Mode point to Business Profiles (SE Ranking).
  3. Publish one reference-style page and one YouTube explainer per core treatment, since AI Mode over-cites reference and health sites and Ahrefs finds YouTube leads AI Overview citations.
  4. Run a 20-prompt patient-question panel on every engine, three times each, and log domain presence weekly.

Open Questions

  • No source measures citation stability or source mix for Perplexity or the Gemini app. Digital Applied promised a Q3 follow-up on ChatGPT Search and Perplexity; it has not been located.
  • Health and dental citation mix is unmeasured: Similarweb broke out beauty, travel and finance only.
  • The AI Mode top-domain conflict (SE Ranking vs Similarweb) is unresolved, and SE Ranking’s market is unstated.
  • QuickSEO’s per-engine MAU and referral figures lack per-figure sources, and AirOps’ collection dates are not stated.

Graph View

Backlinks