Snap put a $2,195 price tag on its newest AR glasses yesterday. KTLA reported it as the opening shot in a new tech battle with Meta. The same news cycle included ENGO Eyewear raising €5.1M for AR sports glasses, Applied Materials joining EssilorLuxottica on AR lens production, and CBS Atlanta announcing a new AR/VR news operation.
Skip the hardware angle. Follow the answers.
Glasses like these do not exist to show a weather widget. They exist to put an answer in front of your eyes while you stand in front of a building, a product, an intersection or a patient. That makes this week's news a search story. Which AI answer engine gets to be the default voice and display layer? And should an SEO team that optimizes for Google page-one links care if that answer comes from Gemini, ChatGPT Search, Perplexity or Copilot?
I ran that comparison today so you do not have to.
What happened this week
The AR glasses news hit in clusters. Snap's $2,195 device is the flashiest item. but it is far from the only signal that the industry is moving beyond smartphone screens.
These are not toy announcements. They are infrastructure signals. Broadcast studios, city tourism departments, sports eyewear companies and luxury lens makers are all treating AR as a normal distribution channel.
The only thing missing from every announcement is a clear answer engine strategy. Nobody wants to buy a $2,195 display and then look at a wall of links.
Why I compared AI answer engines instead of glass specs
I have been testing AI answer engines since the first wave of AI Overviews appeared in Google Search. My usual workflow is to compare them on a desktop browser with clear screens and unlimited reading time. That is not how AR works.
An AR device is a voice-first, context-first interface. The user is standing on a sidewalk. not sitting at a desk. The answer has to be short enough to fit in a small optical field, accurate enough to act on, and fast enough not to feel like a buffering video.
That changes the scoring rubric. So I set up five tasks that mirror real AR usage:
1. Identify a building in front of the user and explain its history.
2. Find the closest coffee shop that is open now and within walking distance.
3. Turn a photo into a reminder with the date and location.
4. Answer a follow-up question without requiring the user to repeat context.
5. Give source names, not just claims, when asked for proof.
I ran those tasks against Google Gemini with AI Overviews, ChatGPT Search, Perplexity, and Microsoft Copilot. I used voice output where possible and enforced a strict response limit of about four sentences.
What the comparison showed
Gemini / Google AI Overviews
Gemini has the strongest local grounding by a clear margin. Google Maps, local reviews, opening hours and business data feed into the answer engine in a way that the others cannot match. When I asked for a coffee shop within walking distance and open now, Gemini gave the most usable answer with the least extra chatter.
Its weakness is presentation. Google AI Overviews are built for a page with links. The same answer on a small AR display feels padded. Too many sentences start with 'based on' or 'according to'. In an audio-only or lens-only format, this slows the user down.
ChatGPT Search
ChatGPT Search wins on conversation memory. The AR user will ask about a building, then a restaurant, then the best route back to the hotel. ChatGPT is the only engine in this group that consistently remembers earlier parts of the conversation without the user restarting the context.
Its weak spot is local data. ChatGPT Search is better at synthesizing knowledge than it is at knowing whether a coffee shop is still open. For a wearable answer engine, that is a serious accuracy gap.
Perplexity
Perplexity is still the best engine for citation transparency. It names its sources. links them, and makes it easy to verify a claim. That is a huge advantage for news, research and trust-building.
On AR, that becomes a problem. Perplexity composes long. paragraph-heavy answers. Voice delivery sounds like someone reading a report aloud. I found myself asking it to 'be shorter' almost every time. A display device cannot render a dense page of citations elegantly.
Microsoft Copilot
Copilot is the middle of the pack. It is competent, reasonably fast and well integrated with Microsoft's data . It does not lead on local search, does not lead on citation depth, and does not lead on conversational memory. It wins when the context is inside an enterprise document environment, not on a street corner.
Quick AI answer engine comparison
| Engine | Best at | Weakest at | AR verdict |
|---|---|---|---|
| Google Gemini / AI Overviews | Local search, maps, place data | Concise voice/display output | Best ready right now |
| ChatGPT Search | Multi-turn context, reasoning | Local accuracy, source density | Best companion for long sessions |
| Perplexity | Citations, transparent sources | Long-form answers | Best for research, worst for glasses |
| Microsoft Copilot | Enterprise data, cross-app workflow | Consumer local knowledge | Dark horse for work-grade AR |
What this means for GEO and SEO
Take the word 'search' out of search optimization for a moment. An AR answer engine is not a search box. It is an assistant that happens to answer questions. The query may be a camera view, not a typed phrase. The user may never learn which website supplied the answer.
This makes the current GEO conversation more urgent, not less. Google's AI Overviews already reshaped the SERP into a summary, and that shift is documented in the New SERP Reality breakdown. AR takes that same logic and removes clicks entirely. A brand that is not named inside the answer is invisible. no matter how high its domain authority is.
That is the same gap my Citation Gap Guide keeps pointing at. You can rank on page one for a keyword and still be missing from the synthesized answer because the answer engine draws from a different source pool. In AR, there is no page two. There is no 'see more' option. There is only the single response that gets displayed.
AR glasses also force websites to answer conversational and local queries properly. A static page about company history will not help when the user asks, 'What am I looking at?' The content has to be structured around entities, places, people and events. That is exactly the type of source data that Gemini and Perplexity are trained to retrieve.
Why agents are the next layer
An answer engine is a passive responder. An AI agent is an active doer. The next AR devices will not stop at answering 'where is the closest open coffee shop'. They will ask whether they should buy you a coffee, then do it after you say yes.
Snap and Meta are already building their glasses around assistant-style features. Google has been moving Gemini from search and information retrieval toward task execution. That is the transition I covered in the AI Agent Reality Check: retrieval alone stops being valuable once the user can delegate the action.
For publishers, the threat is not reduced traffic. The threat is irrelevance. If an AR assistant can book a table, order a product, and ask the vendor to confirm delivery by checking a stored calendar, the website is no longer part of the workflow. The assistant becomes the storefront, the travel agent and the customer service desk.
The brands that survive this transition will be the ones that publish machine-readable facts about availability, pricing and policies. Everything else will be paraphrased away.
What happens next
Expect the next 12 months to be split into two races.
The first race is hardware. Applied Materials joining EssilorLuxottica on AR lens manufacturing is a supply chain milestone, not a product launch. It signals that the optics will get cheaper, lighter and more likely to appear in prescription glasses rather than oversized goggles.
The second race is answer engine default. Meta wants its own assistant on its glasses. Snap has already used OpenAI models for its consumer AI features. so ChatGPT Search has a natural path into those devices. Google has Android. Maps and a distribution network that no one else can match. Microsoft will follow when enterprise AR and Windows devices converge.
This is not a single-winner scenario. But it is also not a democratized marketplace. The key decision is which assistant sits closest to the camera and the microphone.
Takeaway
Do not optimize for the browser tab that no one opens. Optimize for the answer engine that sits on someone's face.
That means structured data, clear entities, accurate operating hours, factual claims with named sources, and content designed to be spoken rather than read. The same work that helps with GEO today makes a brand visible in an AR lens tomorrow.
The $2,195 price tag will drop. The answer engine question will not go away.
Frequently Asked Questions
Which AI answer engine will power smart glasses?
There is no default answer engine yet. Google Gemini, OpenAI ChatGPT Search, Meta AI and Microsoft Copilot are all positioned to reach glasses through different hardware partners and operating systems.
Is GEO better than SEO for AI search?
GEO is not a replacement for SEO. It is the layer that helps a webpage become a cited source inside AI-generated answers instead of just a ranked link in a traditional results list.
Can AR glasses block Google Analytics from tracking visits?
Yes. An AR answer interface can eliminate the click entirely, which means no page visit and no analytics event. That is why brands need to be mentioned inside the answer itself.
Are AI answer engines accurate for local searches?
Gemini has the strongest local data right now because it draws from Google Maps. ChatGPT Search and Perplexity are improving but still less reliable for questions about opening hours. addresses and real-world availability.
Do AI answer engines cite news sources?
Perplexity is the most transparent about citations. ChatGPT Search and Google AI Overviews also include sources. but they are not always specific enough for users who want to verify a claim without reading a full page.
References
> Spent three days on this post. Ran the numbers four times. Exhausting.