I run a small shop called Yellow Pencil out of southern Ontario. We build sites, we fix sites, we argue with clients about H1 tags. Over the last couple of years the thing I get asked most is some version of "am I showing up in ChatGPT yet?" Twelve months ago that was a reasonable-but-fuzzy question. Today it's a specific one with a specific answer, and that shift is most of what this post is about.
I also built RankingLocal.ai because I got tired of guessing. It tracks seven engines now — ChatGPT, Perplexity, Claude, Google AI Overviews, Grok, DeepSeek and Kimi — and I run every Yellow Pencil client through it. So everything below is based on what I'm watching day-to-day on real dashboards for real plumbers, lawyers, dentists, and one guy who sells reclaimed barn wood. It is not based on conference slides.
If you read one thing, read this: the channel mix for local discovery in 2026 looks nothing like it did in April 2025. Five of the seven engines in my dashboard either did not meaningfully cite local businesses a year ago, or did not exist as a real referral source. Now they do. Here is what actually changed.
ChatGPT finally browses, and it cites you
The biggest single shift is SearchGPT being baked into default ChatGPT behaviour for anything remotely local. A year ago, asking ChatGPT "best physiotherapist in Burlington" got you a polite refusal and a suggestion to check Google. Today it returns a real-time browse with 4 to 7 citations, and those citations link out.
For my clients, this shows up as a new traffic source in GA4 labelled roughly as chatgpt.com referrals. It's not huge in absolute numbers — I'm seeing 30 to 300 sessions per month for small local sites — but the conversion rate is ugly good. One dental client is sitting at 11% form-fill rate on ChatGPT referrals versus 2.1% on organic Google. The traffic is smaller but warmer, because the user already got a recommendation and is clicking through to verify.
The practical takeaway: if ChatGPT can't crawl you, you don't exist in this channel. That means no aggressive Cloudflare bot rules, a clean robots.txt that allows OAI-SearchBot and ChatGPT-User, and server-side rendered content. A surprising number of clients came to me in Q1 2026 with Cloudflare "AI scraper" blocking turned on by default, which was silently killing their SearchGPT visibility.
Google AI Overviews stopped being experimental
A year ago AIO was a rolling beta. You could turn it off. You could pretend it wasn't there. In 2026 it is the default experience for most local queries in the US and Canada, and the blue-link SERP you remember is now below a fairly dense AI summary on most "near me" style searches.
What this means operationally: the top three blue-link ranking is still valuable, but it has been demoted. The thing that drives clicks now is being the business that Google chose to name and link inside the Overview itself. That is not the same as being rank 1. Some of my rank-5 clients show up in AIO and some of my rank-1 clients don't, and the delta seems to come down to schema, review velocity, and how cleanly the site answers the specific question Google is trying to summarize.
The one that keeps surprising me is FAQPage schema. I thought it was done when Google pulled the rich result. It's not done. AIO leans on it hard when deciding which source to cite for a "does X offer Y" question. I'm now putting 4 to 6 FAQ items on every service page as a matter of course, and LocalBusiness schema with full areaServed arrays. It is unglamorous 2019 SEO and it still works.
Perplexity went long with Deep Research
Perplexity's Deep Research mode — which launched properly in late 2025 — changed what a Perplexity citation looks like. Instead of a 200-word answer with 3 sources, you get a 1,500-word comparative report citing 15 to 25 sources. For local, this matters because comparative queries ("best X in Y" or "X vs Y") now pull from a much wider set of review aggregators, local directories, and, critically, local business websites that have genuinely useful comparison content.
If you have a page that honestly compares your service to alternatives — not a sales page, an actual comparison — Deep Research will cite it. I have a roofing client whose "metal vs asphalt for Ontario winters" page is now their highest AI-referral earner across any engine. It reads like an honest contractor explaining trade-offs, because that's what it is.
If you only do one thing this quarter, add FAQPage and LocalBusiness schema to your top three service pages, and write one honest comparison article. That combination is still the highest-ROI AI visibility work I can prescribe in 2026, and it costs nothing but a weekend.
Grok got real, at least on X-sourced queries
I was skeptical of Grok for most of 2025. In 2026 I am less skeptical, specifically for anything where X (Twitter) chatter is a real signal — restaurants, events, nightlife, local politics, and any business with active X customer service. Grok's real-time X integration means a restaurant with a steady drip of tagged posts shows up in Grok local queries in a way they don't in ChatGPT or Claude.
It's still a smaller channel than the others in raw volume, but it's the only one where posting on X twice a week measurably moves the needle. For B2B and professional services I'd still deprioritize it. For anything consumer-facing and neighbourhood-level, it's worth 20 minutes a week.
DeepSeek and Kimi matter more than people think
This one is regional and nobody talks about it. I work with clients in Markham and Richmond Hill — cities with very large Mandarin and Cantonese speaking populations — and DeepSeek and Kimi are now genuine referral sources for them. Not hypothetically. I have a Markham accountant getting 40 to 80 monthly sessions from DeepSeek citations, most of them converting.
If your client base is diaspora-heavy, the Chinese AI engines are a channel. If it isn't, you can safely ignore them for another year or two. The monitoring cost is near zero so I keep them in the dashboard regardless, but I don't write bilingual content unless the audience actually warrants it.
What stayed the same
Plenty. Schema is still underrated. Google Business Profile is still the single highest-leverage asset for any local business with a physical address. Review velocity still matters more than review count. NAP consistency across directories still breaks rankings when you move offices. Page speed is still a silent killer on mobile.
The fundamentals did not change. What changed is that the fundamentals now feed seven engines instead of one, and the ones who invested in clean structured data in 2023 are now getting paid on it across all seven.
What I expect over the next 12 months
I try not to predict and I'll keep it short. Three things I'd bet modestly on, not confidently:
First, Google will continue to compress the blue-link SERP, and AI Overview citation placement will become the thing agencies actually sell. Second, at least one of the current seven engines will either merge, shut down, or get acquired — the economics of running a frontier model as a standalone search product are brutal. Third, schema will get a minor but real update specifically for AI-citation hints, probably driven by Schema.org rather than a single vendor.
Beyond that, I don't know. Anyone who tells you they do is selling something.
If you want to see what your business looks like across these seven engines today, I built a free checker at /free-tools/ai-visibility/. It'll tell you which engines cite you, which don't, and what the gap looks like. No signup, no upsell screen, just the numbers.
Questions, arguments, or your own data that contradicts mine — hello@rankinglocal.ai is read by me directly.