Key takeaways
- Meta AI, Alexa+ and Siri each pick brands through a different mechanism: web crawling, Amazon's catalog and ads, and on-device personal context. One GEO playbook will not cover all three.
- Meta is building its own web index. Meta-WebIndexer reached 37.8% of tracked AI crawler requests on Aug 9, 2026, up from about 2.2% in mid-July (Promptwatch Data).
- Alexa+ brand visibility currently runs through Amazon's catalog and Sponsored Brands. Branded Conversations, announced Sep 29, 2026, is due in the US in January 2027.
- Siri's rebuilt version is still in beta. There is no public data on how it chooses brands, so be skeptical of anyone claiming otherwise.
- Nobody has first-party citation data for these three assistants. Web-grounded assistants like ChatGPT, Perplexity and AI Overviews are the best available proxy.
Why personal assistants are a different problem
Most AI visibility advice is written about ChatGPT, Perplexity and Google AI Overviews. Those are search-style products. You type a question, get an answer with a list of sources, and you can measure which domains got cited.
Personal assistants work differently. Meta AI lives inside WhatsApp, Instagram and Messenger. Alexa+ lives in the kitchen and on the shopping screen. Siri lives on the phone and increasingly reads your own mail and messages. The answer is shaped by who is asking, where they are, and what the platform wants to sell or surface.
That makes brand visibility messier. I'd rather say that upfront than pretend there is a clean ranking factor list. There isn't one for any of these three, and that is the most honest starting point.
The three assistants at a glance
| Assistant | Where it lives | Pricing | How brands get surfaced | Evidence quality |
|---|---|---|---|---|
| Meta AI | WhatsApp, Instagram, Messenger, Ray-Ban glasses, meta.ai | Free | Web crawling (Meta-WebIndexer), Marketplace, Meta surfaces | Partial: crawler data, secondary sources |
| Alexa+ | Echo devices, Alexa for Shopping | Included with Prime ($14.99/mo or $139/yr) or $19.99/mo standalone | Amazon catalog, Sponsored Brands, Branded Conversations | Good for ads, thin for organic |
| Siri (rebuilt) | iPhone and Apple devices | Free on supported devices | Personal context, in-app actions | Almost nothing public |
Pricing and positioning come from the Agent37 roundup of personal AI assistants. Alexa+ runs on Amazon Nova plus Anthropic models, per the same source.
Meta AI: the one with a visible crawl footprint
Meta AI is the assistant where brands have the most to act on right now, mostly because Meta has been loud about crawling.
What the crawler data shows
Promptwatch Data tracked Meta-WebIndexer at about 2.2% of AI crawler requests in mid-July 2026. By Aug 9 it hit 37.8%, roughly a 17x rise in under a month, making it the heaviest AI crawler in Promptwatch's logs. The share sat near 2% until Jul 17, spiked to about 23% on Jul 20 to 22, then surged again from Aug 5. You can see the daily chart in Promptwatch's Meta-WebIndexer report.
Two caveats matter. The numbers cover the meta-webindexer user agent only, and they describe crawling, not citations. Promptwatch's read is that Meta is building a search index so Meta AI doesn't depend on Google or Bing. That is an inference, not something Meta has confirmed. It does line up with older reporting: The Information said in October 2024 that Meta was building an AI search engine, and at the time Google and Bing supplied Meta AI's news, stock and sports answers.
What Meta says to do
Secondary sources relaying Meta's crawler documentation say that allowing Meta-WebIndexer in robots.txt helps Meta cite and link to your content in Meta AI responses. The suggested rule is:
User-agent: Meta-WebIndexer
Allow: /
I could not verify this against Meta's own page, so check the official crawler docs before you ship it.
Promptwatch also suggests auditing robots.txt and server logs for Meta-WebIndexer, Meta-ExternalAgent, Meta-ExternalFetcher and FacebookBot, and budgeting for sustained crawl load. One developer, @levelsio, reported load alerts from Meta scraping in early August. If you run a small server, that is worth taking seriously.
There is a timing angle too. Promptwatch notes that whether your content was crawled during the build-out may decide whether it can be cited at launch. That is speculation, but cheap to hedge against: letting the bot in costs nothing.
What Meta AI recommends
According to a secondary guide, Meta AI runs on Muse Spark 1.1 in the Meta AI app and meta.ai. It can research across the web and produce cited reports, surface products from the web and Marketplace, and give local recommendations. Another source says Muse Spark became the model behind Meta AI in April 2026. Both are secondhand, so treat them as directional.
Two more points from the same guide are worth remembering:
- Facebook AI Mode can ground answers in public content from Groups, Reels and other Meta surfaces.
- Meta has expanded the data used to personalize responses, so two users asking the same question can see different brands.
Meta has not published a citation ranking formula. Schema markup, entity clarity and similar tactics are reasonable GEO practice, but they are not confirmed signals. Anyone selling you a "Meta AI ranking checklist" is guessing.
Social content is a mixed bet
One thing to be careful with: Meta owns Facebook and Instagram, but that doesn't mean social posts dominate AI answers elsewhere. In Promptwatch Data for Sep 28 to Oct 4, 2026, social platforms were only 0.46% of ChatGPT citations, 2.7% of Perplexity citations and 12.2% of Google AI Overviews citations. Facebook was 2.38% in AI Overviews, more than double its earlier share. The social media citations report has the full breakdown. It says nothing direct about Meta AI, which is the gap.
Alexa+: visibility is a retail and ads problem
Alexa+ is conversational now. It holds context, reads photos and documents, and orders food. Reviews still cite factual errors, device lag and a buggy app, per the Agent37 roundup. For brands, though, the interesting part is shopping.
Branded Conversations
On Sep 29, 2026, Amazon Ads announced Branded Conversations, due in the US in January 2027. It turns Sponsored Brands prompts into multi-turn conversations inside Alexa for Shopping. The details from Amazon's announcement:
- It draws on the advertiser's catalog and brand-supplied content such as product guides, compatibility info and usage tips.
- Answers are personalized to shopper preferences and purchase history.
- Eligible Sponsored Brands collections campaigns are enrolled automatically. Authors are excluded, and it is open to US endemic advertisers at first.
- Billing is CPC on the conversation-starter prompts only. Follow-up turns are free.
- Brands can supply a campaign-level brief covering differentiators, launches, product-to-need mapping or routines.
- You can pause individual prompts, opt out at campaign or account level, and see reporting in the Ads Console or API. The prompts tab only shows prompts after they get a click.
Press coverage, including eMarketer, frames this as brands getting more influence over how they appear in Alexa responses. I'd put it more bluntly: the mechanism for being recommended by Alexa while shopping is increasingly pay-to-play, with organic catalog quality underneath.
What to do before January
- Clean up your Amazon product detail pages. Compatibility info, usage tips and differentiators are exactly what the brief asks for, so write them once and reuse them.
- Run Sponsored Brands collections campaigns now, so you have click history when the conversation format arrives.
- Write the campaign brief like a sales enablement doc: who the product is for, what it replaces, what it doesn't do.
This is not web citation. Don't expect GEO work on your own site to move Alexa's shopping answers much. Amazon's catalog is the source of truth there.
Siri: the honest answer is we don't know yet
Apple's rebuilt Siri was shown at WWDC in June 2026 and ships "in beta later this year", per TechCrunch's coverage as summarized by Agent37. It arrived roughly two years late, with real personal context and in-app actions. The Siri on most phones today is still the old one.
I found no primary-source data on how Siri picks or recommends brands. Apple hasn't published citation or ranking behavior for the new version, and no tracking vendor has credible data on it yet. So the sensible stance is to prepare for the things that are likely to matter without claiming to know the mechanism:
- App Intents and in-app actions. If Siri completes tasks inside apps, brands with well-structured app integrations are likelier to be the thing it reaches for.
- Your Apple Maps and Business Connect listing, if you are a local business. This is a reasonable bet, not a confirmed factor.
- Plain web fundamentals: crawlable pages, accurate facts, consistent brand information. These help every system.
If you read a confident "how to rank in Siri" guide in 2026, check whether it cites anything. Most don't.
What the web-grounded data tells us
Since first-party data on the three assistants doesn't exist, the best proxy is how web-grounded assistants cite sources. A few findings from Promptwatch Data that hold up regardless:
- Product pages were the most cited content type in ChatGPT at 22.4% of citations (Sep 28 to Oct 4, 2026), followed by landing pages and listicles. See the content type report.
- Landing pages and docs stayed cited longest, while news lost half its pages within 36 days, per the content half-life report.
- Almost half of ChatGPT responses cite 10 or more unique domains in the unique domains report.
Source counts also vary by platform. A Promptwatch page dated Sep 23, 2025 put ChatGPT at about 5 sources per web-search answer, AI Overviews at about 10, and Perplexity at almost exactly 10 (average sources per response). That page is old, so check it before quoting numbers.
The practical takeaway: clear product and landing pages that state facts plainly are useful everywhere. Meta AI's web index and Alexa's catalog both reward the same boring basics.
A practical plan for 2026
| Step | Meta AI | Alexa+ | Siri |
|---|---|---|---|
| Technical | Allow Meta-WebIndexer, review other Meta bots, watch server load | Complete catalog data and detail pages | Clean app intents, accurate local listings |
| Content | Plain product and landing pages, public social presence | Product guides, compatibility and usage content | Consistent brand facts across the web |
| Paid | Not applicable yet | Sponsored Brands now, Branded Conversations in Jan 2027 | None known |
| Measure | Server logs, manual prompt tests | Ads Console reporting | Manual spot checks only |
Step 1: check your logs
Open your server or CDN logs and search for Meta-WebIndexer, Meta-ExternalAgent, Meta-ExternalFetcher and FacebookBot. If you see nothing, your robots.txt or firewall may be blocking them. If you see a flood, rate-limit before it hurts your hosting bill.
A tool like Promptwatch can show IP-verified AI crawler activity per page, including Meta's bots, which saves you from parsing raw logs by hand.
Step 2: test prompts by hand
Meta AI answers vary by user, so fixed-prompt tests are a benchmark, not a universal result. Write ten prompts your customers might ask, run them from a few accounts, and screenshot what appears. It is tedious. It is also the only direct measurement available for Meta AI, Alexa+ and Siri today.
Step 3: track the proxies
For web-grounded assistants, monitoring tools help. Promptwatch tracks ChatGPT, Gemini, Perplexity, Google AI Overviews, Meta AI and Amazon Alexa product results through real interfaces, and it also covers shopping carousels and ads in AI search. Among the other options in the catalog are Profound and Otterly.AI, which focus more on monitoring.
Profound
Otterly.AI
For a broader list, browse the GEO software directory at bestgeosoftware.com.
Step 4: fix content that is thin
Find the questions assistants answer about your category and check whether you have a page that answers each. Content gap analysis does this systematically. Otherwise, a spreadsheet of questions and URLs works fine at small scale.
What not to believe
- Claims of specific Meta AI, Alexa+ or Siri citation shares. No one has first-party data on them yet.
- Lists of "ranking factors" for any of the three. Meta hasn't published a selection formula, and Apple and Amazon haven't either.
- The idea that Meta-WebIndexer's crawl share equals citation share. It is a crawler metric.
- Any suggestion that blocking Meta's crawler is harmless. It may be a smart choice for some publishers, but you lose any chance of being cited by Meta AI.
Bottom line
Treat these three assistants as three separate channels. Meta AI is the one to act on today: let the crawler in, watch your logs, and keep your pages factual. Alexa+ is an ads and catalog play, with Branded Conversations arriving in January 2027. Siri is a wait-and-see, so keep your fundamentals clean and revisit once the beta ships and someone publishes real data.
And if you want a baseline now, run the manual prompt test this week. It takes an afternoon and will tell you more than any guide, including this one.