The eight AI engines, in plain English.

Why eight, and not one

The phrase "AI search" usually shows up in business press as if there is a single new system replacing Google. There is not. There are at least eight production assistants that route real informational traffic in 2026, and a handful more on the way. Each one has a different user base, a different retrieval bias, and a different relationship with the open web. The same local business can be the first answer on Perplexity, missing entirely from Meta AI, and described as the wrong company by Grok. We have seen all three happen in a single audit.

That variation is the entire reason we audit against all eight. A score on one engine is a vanity number. The signal that matters is the spread.

The eight, one at a time

We will go in rough order of installed user base and likely impact on local business traffic over the next 24 months, not in alphabetical order. Numbers below are the working-model estimates we use internally as of May 2026, drawn from the public usage data we trust most and tempered by what we actually see retrieving in field audits. Treat them as directional, not gospel.

1. ChatGPT (OpenAI)

The default name. By any sensible measure of weekly active users among consumer AI assistants, ChatGPT is the largest. The under-40 demographic in the United States treats it the way the over-50 demographic treats Google. The product also got search built into it in late 2024, which means a ChatGPT user asking "best restaurant for an anniversary in Portsmouth NH" is now a customer doing local discovery the same way a Google search was local discovery a decade ago. Most owners we audit do not appreciate how aggressively ChatGPT has shifted into real-time retrieval.

Who uses it for a Seacoast business search: Anyone under 50 with a phone, a laptop, and a question. Increasingly common as a first step for restaurant, lodging, and home-service decisions where the user does not already have a name in mind.

How it retrieves a local business: A hybrid of long-trained corpus knowledge plus real-time browsing. If your site is structured for retrieval, ChatGPT in search mode often cites you directly with a clickable source link. If your site is not structured, ChatGPT pulls from third-party aggregators (Yelp, TripAdvisor, the local chamber) and you become a paraphrase rather than a citation.

2. Google AI Overviews

The quietest of the eight, by design. Google AI Overviews is the generated paragraph that now appears above the blue links on a meaningful share of US search results. Most users do not realize they are reading an AI summary. They read it, decide whether they got their answer, and either click a citation or close the tab. The traffic loss to local businesses from Overviews has been the single biggest story in SEO for the last 18 months. It is also the story that gets buried fastest because it is harder to measure than a direct ChatGPT search.

Who uses it for a Seacoast business search: Everybody who still types into Google. Which is, still, most people. The Overview sits in the path between the query and the result.

How it retrieves a local business: Same crawl as classic Google, but a fundamentally different summarization step. Sites that rank well in classic Google sometimes do not appear in the Overview. Sites that are weak in classic SEO sometimes do, because the Overview rewards structured, scannable, entity-clear content over keyword density. The two systems do not always agree about who the right answer is.

3. Google Gemini

The standalone assistant. Gemini is built into Pixel phones, Android, Gmail, Docs, Calendar, and the Google app on iOS. Its installed footprint is enormous and silent. The same user who types into Google during the day is increasingly asking Gemini through their phone microphone in the evening. We see Gemini retrieve very differently from Overviews, despite both being Google products. Gemini reaches further into the open web, weighs schema markup more aggressively, and is more willing to summarize a long page than the Overviews layer is.

Who uses it for a Seacoast business search: Pixel and Android owners; Gmail users with the assistant turned on; anyone who has tapped the Gemini button without realizing they were not using regular Google.

How it retrieves a local business: Live web fetches, heavy weighting of structured data, and a noticeable preference for sites where the entity definition (what the business actually is) is stated plainly in the first 200 words. Sites built for image-first storytelling tend to underperform.

4. Microsoft Copilot

Built into Windows 11 directly, integrated into Edge, integrated into Outlook, integrated into Bing search. Copilot reaches an audience that ChatGPT does not: the office-using, 45-to-65, professional-services-buying customer who is more likely to make a high-ticket local purchase than the average ChatGPT power user. We do not audit Copilot because it is large. We audit it because the customer behind a Copilot answer about a CPA, a wealth manager, or a plastic surgeon is usually a higher-value lead than a ChatGPT answer about the same category.

Who uses it for a Seacoast business search: Windows-laptop professionals, Edge browser users, Outlook users with the Copilot panel open. Heavy on financial services, legal, accounting, and medical specialty searches.

How it retrieves a local business: Bing index plus live retrieval, with a strong preference for sites with clean Schema.org markup and verifiable third-party signals (BBB, professional licensing boards, industry associations). Copilot is the engine most likely to surface a regulated-profession site if that site has plain professional credentials on it.

5. Anthropic Claude

Different audience. Claude has the smallest consumer footprint of the major five but a disproportionate use in professional services, legal, technical, and analytical work. It is the assistant most likely to be asked a careful comparison question rather than a quick recommendation question. A buyer asking Claude "compare three estate-planning attorneys in the Seacoast on these five criteria" is doing higher-intent research than a buyer asking ChatGPT the same question casually.

Who uses it for a Seacoast business search: Professional buyers, knowledge workers, anyone whose company has standardized on Claude for internal work and now uses it for personal decisions out of habit.

How it retrieves a local business: Conservative. Claude is the engine most likely to decline to name a specific business if it is not confident. That sounds neutral until you realize: if your business is the one the engine declined to name, you lost that buyer. Claude rewards sites where the entity definition is unambiguous, the location is unambiguous, and the third-party signal set is consistent across the open web.

6. Perplexity

The citation-first engine. Perplexity is the assistant most likely to attach explicit sources to every claim in its answer. That is its product positioning: not "I am smart," but "I will show you the evidence." For owners, this matters more than it sounds. A Perplexity answer that mentions your business in the prose but cites a Yelp page as the source has just sent that buyer to Yelp, not to you. Perplexity rewards businesses that are the cleanest, most-structured, most-trustworthy source on a topic. It is the engine where good GEO work pays back fastest.

Who uses it for a Seacoast business search: Buyers who want to verify. Heavy use among high-spend prospects (legal, financial, medical, luxury hospitality). Smaller user base than ChatGPT, higher per-user conversion rate to action.

How it retrieves a local business: Real-time web retrieval, citation-first ranking. Sites with consistent name and address and clear authoritative content tend to become the cited source. Sites that read like marketing copy tend to be paraphrased but uncited.

7. Meta AI

The engine you forget about. Meta AI is inside Instagram, WhatsApp, and Facebook. Most of its users do not realize they are using an AI assistant at all. They tap the search bar in Instagram, type "best brunch in Portsmouth," and read the answer Meta AI generates above the regular results. This reaches a demographic that does not type URLs into a browser at all. For local hospitality, retail, and consumer-service businesses in the Seacoast, ignoring Meta AI is a quiet way to lose the Instagram-native customer.

Who uses it for a Seacoast business search: Instagram-first buyers, WhatsApp users, the under-35 demographic doing visual discovery before they pick a place.

How it retrieves a local business: A blend of Meta's own social signals (your Instagram presence, your Facebook page, tagged photos) with open-web retrieval. The engine that most rewards businesses that have invested in their owned social channels in addition to their website.

8. Grok (xAI)

Inside X (formerly Twitter), and the most idiosyncratic of the eight. Grok has a smaller user base than any of the other seven, but a meaningfully different retrieval bias. It pulls heavily from X itself (so businesses with active X presence get cited there in ways they get nowhere else) and from a different slice of the open web than its competitors. We have seen Grok return citations for a business that no other engine surfaces, and miss businesses that every other engine retrieves cleanly. It is the outlier that prevents single-engine optimization from being good enough.

Who uses it for a Seacoast business search: Active X users; tech-leaning buyers; the buyer who treats AI assistants as an information sport.

How it retrieves a local business: X-platform signals weighted heavily, open-web retrieval as a secondary layer. Most local businesses in the Seacoast are completely absent from Grok citations because they have no meaningful X presence and have not been picked up by anyone who does.

Why the spread between engines matters

If you only had to be good at one of these, the GEO problem would be a sub-problem of regular SEO. The bad news for owners is that the engines do not agree. The good news is that the disagreements follow patterns, and the patterns are addressable. We see roughly the following:

• Schema.org markup moves you on Gemini, Google AI Overviews, Copilot, and Perplexity.
• Clean entity definition (plain first-paragraph statement of what the business is) moves you on Claude, ChatGPT, and Gemini.
• Consistent third-party signals (matching name, address, phone across Google Business, the chamber, your industry associations) move you on Copilot, Claude, and Perplexity.
• Active social channels move you on Meta AI and Grok in ways that no website change alone can.
• Crawlable text rather than image-first storytelling moves you on all eight.

Most of the audit work we do is not about one engine. It is about closing the spread. A business that is fully retrieved on three engines and invisible on five does not need a new website. It needs the structural work that brings the bottom five up to where the top three already are.

Which engines will quietly drive most of your traffic by 2027

We do not believe anyone can name the winners cleanly. The honest version, from the field:

• ChatGPT plus Google AI Overviews together account for the majority of AI-routed visits in the United States today, and we expect that share to stay above 50 percent through 2027.
• Gemini and Copilot are the silent giants. They reach customers who do not realize they are using an AI. We expect both to surpass Perplexity in routed local traffic by the end of 2026.
• Perplexity is the engine where good GEO work pays back fastest in absolute conversion rate, even if its raw user base stays smaller than the giants.
• Meta AI is the engine the visual-discovery hospitality and retail brands cannot afford to be invisible on.
• Claude is the engine to watch for high-LTV professional services. Disproportionate impact per query.
• Grok is the wild card. Worth being visible on. Not worth optimizing for first.

The strategic move for a Seacoast business is not to chase the top one or two. It is to close the spread. Get retrieved on all eight to a baseline level. Then improve the engines where your category over-indexes (Perplexity for legal, Meta AI for hospitality, Copilot for financial services, and so on).

What an owner should do this week

Open all eight in browser tabs if you can. Ask each one the question your best customer would ask if they did not already know you. Do not name yourself. Note three things:

• Which engines named you.
• Which engines named you accurately.
• Which engines cited you, and which paraphrased you while citing somebody else.

That table is the working baseline for everything we do next. If you would like us to run that table for you across all eight engines and five queries, we do that for free for any Seacoast business that asks. It is the same starting baseline we use on a paid engagement, so the report tells you what to fix even if you do not hire us to fix it.

For more on the Portsmouth-specific field data behind the engines, see how a Portsmouth business shows up in AI search (or does not). For the professional-services slice (dental, legal, accounting, medical aesthetics) see how customers find a dentist, lawyer, or CPA through AI search.