ChatGPT vs Perplexity vs Gemini: Where to Focus

You have one budget and three logos shouting at you. Every vendor deck says you need an AI visibility strategy for all of them. Most of those decks never tell you the size of the prize on each engine, because the honest numbers are unflattering.

So here they are. This post compares ChatGPT, Perplexity, and Gemini on the four axes that actually decide where a service business spends: how each one retrieves sources, what it cites, how much traffic it realistically sends you, and how hard it is to influence when your domain has no authority yet.

Which AI engine should a service business optimize for first?

None of them. Optimize Google's classic index first, because that is still where 86.67% of US searches happen (StatCounter, June 2026) and it is the substrate every other answer is built on. Then ChatGPT, which took 71.33% of US AI referral traffic in 2026. Then Gemini at 12.87%. Perplexity, at 6.85% and falling, is the last thing on the list.

Those AI shares come from SE Ranking's analysis of 101,574 websites with Google Analytics installed, covering January 2025 through April 2026. It is the largest public referral dataset we could verify, and it puts a hard number on a soft argument.

The uncomfortable part is the denominator. In the same study, AI platforms sent 0.29% of all US website traffic. Organic search sent 42.75% globally. SE Ranking's own summary: Google still sends 134x more visitors than every AI platform combined. Perplexity's 6.85% is a slice of a slice.

How does each engine actually retrieve and cite sources?

They use three different retrieval systems, which is why the same page gets cited by one and ignored by another. Perplexity runs its own crawler and its own index. ChatGPT runs a separate search crawler from its training crawler. Gemini and Google's AI features sit on top of Google's index and use query fan-out.

OpenAI's crawler documentation splits the jobs explicitly. OAI-SearchBot is the one that surfaces you in ChatGPT search answers — OpenAI writes that sites opted out of OAI-SearchBot 'will not be shown in ChatGPT search answers.' GPTBot is for training data only. ChatGPT-User fetches a page when a user's prompt triggers a visit, and OpenAI notes it 'is not used for crawling the web in an automatic fashion,' so robots.txt rules may not apply to it.

That distinction matters more than any content tactic. Block GPTBot to protect your content from training and you keep your ChatGPT search visibility. Block OAI-SearchBot and you have deleted yourself from ChatGPT's answers. The two get confused constantly, and the mistake is silent.

Perplexity's docs mirror the structure: PerplexityBot is 'designed to surface and link websites in search results on Perplexity' and is not used for model training, while Perplexity-User handles live user-triggered fetches and 'generally ignores robots.txt rules.' Ahrefs notes Perplexity does not draw on Google's or Bing's index at all — it retrieves from its own.

Google is the outlier in the other direction. Its documentation states there are no additional technical requirements for AI Overviews or AI Mode: a page must be indexed and eligible to show with a snippet, and that is it. Both surfaces use query fan-out — issuing multiple related searches across subtopics — to assemble the answer.

How many sources does each engine cite in a typical answer?

No credible public dataset gives an apples-to-apples average citation count per engine, and we are not going to invent one. The one measured figure that exists is for Google's AI summaries: Pew Research Center tracked 68,879 real Google searches and found 88% of AI summaries cited three or more sources, with only 1% citing a single source. The median summary was 67 words.

The per-engine numbers floating around SEO blogs — 'Perplexity cites 21.87 sources, ChatGPT cites 7.92' — trace back to vendor posts with no published methodology. Treat them the way you would treat a contractor's verbal quote.

What is measured, and far more useful, is which sources each engine reaches for. Profound analyzed 680 million citations from August 2024 to June 2025 and found three genuinely different personalities:

Read the Perplexity row again. Yelp and TripAdvisor both crack its top ten cited domains. For a plumber or a med spa, that is a real instruction: your Perplexity visibility is mostly built on third-party profiles you do not own, not on your blog. Perplexity is the one engine where fixing your Yelp listing beats writing another post.

How much referral traffic does each realistically send a local business?

Roughly one in every 312 website visits now comes from an AI search engine — 0.32% of global traffic in 2026, up from 0.02% in 2024 (SE Ranking, 101,574 websites). In the US it is 0.29%. If your site gets 3,000 visits a month, that is about nine AI-referred visits, and roughly six of them are ChatGPT.

Anyone selling you a five-figure Perplexity program should have to say that out loud. On the US numbers, Perplexity is 6.85% of 0.29%. It is not a channel yet. It is a signal.

The counterweight is engagement. In the same study, visitors arriving from AI platforms spent 67.7% more time on site than organic-search visitors — about 9 minutes 19 seconds versus 5 minutes 33 seconds on average (SE Ranking, engagement data collected January–April 2025). In the US specifically the gap was smaller, at 28.7%. AI traffic is thin and pre-qualified, not thick and cheap.

Direction matters too. ChatGPT's share of AI referrals fell from 79.74% in 2025 to 74.78% globally in 2026 — but its actual traffic still grew 27%. Gemini grew 231% year over year and passed Perplexity in the US in January 2026. Perplexity's US share fell from 11.42% in 2025 to 6.85% in 2026. One of these is compounding and one is decaying.

Which engine is easiest to influence from a zero-authority domain?

Perplexity — by a wide margin — which is exactly why it is a trap. Ahrefs tested 15,000 long-tail prompts and found 28.6% of Perplexity's cited URLs rank in Google's top 10 for the same query, versus roughly 8% for ChatGPT, Gemini, and Copilot. Perplexity is the most rank-aligned engine, so classic SEO moves the needle there fastest.

There is a second reason Perplexity looks winnable. It runs its own index off its own crawler, which is far smaller than Google's. A smaller index means less competition for any given citation slot, and it means a new page can get picked up without waiting on Googlebot's recrawl schedule. A fresh domain can genuinely show up in a Perplexity answer within days.

But easiest-to-move is not the same as worth-moving. You would be optimizing hardest for the engine that sends the least traffic and is losing share. The winnable engine and the valuable engine are not the same engine, and that gap is where most GEO budgets die.

For a fresh domain the realistic order of operations is unglamorous: get indexed, get your entity consistent across Google Business Profile and the review platforms, earn citations on the sites the engines already trust, then publish depth. Our GEO service runs that as one program with SEO rather than as a separate line item, because the retrieval layer for every engine still starts with pages Google can crawl.

Does optimizing for one engine carry over to the others?

Partially, and less than vendors claim. In Ahrefs' 15,000-prompt study, only 12% of citations across ChatGPT, Gemini, and Copilot appeared in Google's top 10 for the original prompt, and overlap with Bing's top 10 averaged about 10%. Ranking #1 does not buy you a citation.

The reason is query fan-out. Ahrefs explains that assistants retrieve against multiple rewritten variations of your prompt and then fuse the results, so a page ranking #6 for three sibling queries can beat a page ranking #1 for one. Google confirms AI Overviews and AI Mode use the same technique to surface 'a wider and more diverse set of helpful links.'

The practical carry-over is real but narrow. Crawlability, textual content, structured data that matches the visible page, and internal links help everywhere — Google says as much in its own AI-features guidance. What does not carry over is the citation itself. Winning Perplexity teaches you nothing about winning ChatGPT, because they read different parts of the internet.

If you want the engine-specific tactics, we wrote them separately: Perplexity's recency play, getting recommended by ChatGPT, and earning citations in Google AI Mode. This post decides which of those three you read first.

Where do Google's AI Overviews and AI Mode fit against the chatbots?

They are bigger than all three chatbots combined for most service businesses, because they live inside the search results your customers already use. Pew Research Center found 18% of real Google searches produced an AI summary back in March 2025. Ahrefs, analyzing 146 million SERPs, put AI Overviews on 20.5% of keywords by September 2025.

That reach comes with a cost you should budget for. Pew tracked the actual browsing of 900 US adults and found that when an AI summary appeared, users clicked a search result on just 8% of visits, versus 15% when no summary appeared. They clicked a link inside the summary itself on 1% of visits. Getting cited does not hand you the click — it hands you the mention.

Google's AI surfaces also behave more like search than the chatbots do. Ahrefs' March 2026 study of 863,000 SERPs found 38% of pages cited in AI Overviews also rank in Google's top 10 for that query — down from about 76% a year earlier, but still triple the ~12% overlap the chatbots show.

And the commercial case is measurable. Seer Interactive analyzed 5.47 million queries across 53 brands and found that being cited inside an AI Overview delivers 120% more organic clicks per impression than sitting on the same SERP uncited — though still 38% fewer clicks than a SERP with no AI Overview at all.

One caveat that saves service businesses money: Ahrefs found only 7.9% of local searches trigger an AI Overview, and only 4.3% of commercial-intent and 2.1% of transactional-intent keywords. Informational queries trigger at 21.4%. If your money keywords are 'emergency plumber near me,' AI Overviews barely touch them. If they are 'how much does a water heater replacement cost,' they dominate.

What is the honest verdict for a $2,500/mo budget?

Spend roughly 80% of your visibility work on Google's index and its AI surfaces, 15% on ChatGPT, and 5% on everything else — and do not run a dedicated Perplexity program at all. Here is the same comparison in one table.

The verdict, stated plainly: Google wins, ChatGPT is second, Perplexity is a rounding error for a local service business. Anyone who tells you otherwise is selling a Perplexity dashboard.

That is not an argument against GEO. It is an argument against buying GEO as four separate products. The work that gets you cited in ChatGPT — crawlable pages, clear entity signals, third-party mentions on sites the models already trust, content that answers a literal question in the first sentence — is the same work that gets you into AI Overviews. Which is why we run SEO, GEO, and AEO as one program in the base retainer instead of billing them as three.

One more piece of housekeeping. Google does not break AI Overview or AI Mode traffic out separately in Search Console — its docs confirm it lands in the overall 'Web' search type. So before you spend a dollar chasing any engine, set up referral tracking that can tell ChatGPT traffic apart from everything else. If you cannot measure it, you cannot kill it, and a channel that produces no qualified leads in 90 days should get killed.

If you want to know which engines are already citing you and which of your pages they are pulling from, that is the first thing we check. Start with a free audit — we will show you your current AI citation footprint and tell you honestly whether it is worth chasing. Get my free audit.