Toronto AI Citation Tracker: August 2026

The short version:

• Across the same 50 questions we ran in July, AI assistants named a specific Toronto or GTA business 97% of the time this month, up from 89% in July. This is edition two, so it is our first real month-over-month read, and the headline direction is up across the board.
• Perplexity had the biggest month of any engine. It jumped from 74% to 92%, an 18-point gain that closes most of the gap it opened in July. Google AI Mode and ChatGPT both hit 100% this month. Claude sits at 96%.
• The general contracting category had the sharpest swing of any category: 75% in July to 95% in August. The clearest example is a bathroom-renovation-budget question that returned a generic cost guide with no business named in July, and this month returned a specific shortlist on three of the four engines.
• Claude's account-personalization problem, which we flagged as a one-off caution in July, got worse this month. On two separate pest-control queries it recognized the founder's own client from account history and answered as a competitive check for that client instead of the neutral consumer question we asked. One of those answers did not address the actual query at all.
• Perplexity's Scarborough problem showed a partial improvement. It still resolved one query to the wrong country in its map data, but for the first time the written answer led with the correct Toronto businesses before mentioning the geographic mix-up, rather than returning UK results outright.
• ChatGPT's ad presence on our 50 hyper-local questions dropped to zero this month, down from 2 in July, more evidence that local-service ad inventory is thin and volatile rather than a real trend in either direction.

Why we run this

Nobody publishes a Toronto-specific measure of who the AI assistants actually recommend. We started this tracker in July to build that record ourselves, one month at a time, with the same 50 questions run the same way so the numbers stay comparable. Edition one gave us a baseline. This is the first month we can say something moved.

How we measured it

Same 50 fixed questions as July, across the same 10 categories: dental, personal injury law, accounting, pest control, plumbing, general contracting, restaurants, wellness clinics, medical clinics, and home services. Every question went to all four assistants, Google AI Mode, Perplexity, ChatGPT, and Claude, giving 200 data points. Collection ran August 8, five days later than the template's first-business-day target; we are stating that gap plainly rather than pretending the cadence held.

Three answers were excluded from the percentages rather than forced into a yes or no: two Perplexity queries returned a persistent site error across multiple retries, and one Claude query is covered in Finding 3 below.

One methodology note we would rather disclose than bury: due to a session-management slip on our end, Google AI Mode's full 50 queries, five of Perplexity's, and all of ChatGPT's ran in a real browser signed into our own accounts rather than a fully anonymous session. We checked the answers themselves for signs of that leaking in, and found none. The businesses named lined up with what the same queries returned on other engines run cleanly, and none of the answers referenced our accounts, our clients, or our history the way Claude's did (see Finding 3). We are disclosing it anyway, because a session slip we did not catch until partway through is worth naming, and we have tightened our process for next month.

Finding 1: Local citation rate jumped 8 points, and every engine improved

Across all 197 scoreable answers this month, an assistant named a specific Toronto or GTA business 97% of the time, up from 89% in July. That is an 8-point gain in a single month, and no engine went backward.

Google AI Mode and ChatGPT now both name a local business on every single one of our 50 questions. Claude is close behind. The real story is Perplexity, which closed most of the 24-point gap that separated it from Google AI Mode in July, in a single month. We cannot yet say whether that is a genuine product improvement or a good month, since we have exactly one prior reading to compare against. That is precisely the question next month's edition answers.

Finding 2: General contracting had the sharpest category swing

Local presence moved in every category except two that were already at 100%. The biggest jump was general contracting, up from 75% to 95%.

The contracting swing has a specific cause we can point to. Our bathroom-renovation-budget question ("bathroom renovation Scarborough budget") returned a generic cost breakdown with no business named on all four engines in July. This month, Google AI Mode, ChatGPT, and Claude all named the same two businesses, a Scarborough renovation firm and a general contractor, while only Perplexity stayed generic on that exact question. A single question flipping across three engines at once reads less like noise and more like the underlying local pack catching up to a cost-shaped query it previously treated as pure research.

Finding 3: Claude's account-personalization problem got worse, not better

We flagged this as a caution in edition one: Claude, the one engine we run signed in, occasionally read our own account context into an answer meant to represent what a stranger would see. This month it happened twice, on consecutive pest-control questions, and the second instance was more serious than anything we saw in July.

On "cockroach treatment Scarborough," Claude opened by framing the answer as a competitive check for one of our own pest-control clients, naming that client and describing where it sits against the businesses it then listed. The consumer-facing local names it returned were legitimate, so we kept the answer, but the framing was shaped by account history no first-time visitor would carry.

On the very next query, "commercial pest control Toronto restaurant," Claude did not answer the consumer question at all. It opened by telling us where our client ranks organically for that exact search term and ran the entire response as an SEO positioning check, never producing a shortlist a restaurant owner asking that question could use. We recorded that answer as refused and excluded it from the percentages, because it is not a comparable answer to the other 199 in this dataset.

Two incidents in one category, one of them a full derailment, is not a one-off anymore. If you are checking your own AI visibility using an AI assistant you are signed into daily, this is exactly the failure mode to watch for: the tool may be telling you what it already knows about your business, not what a stranger asking the same question would see. Log out, or ask from an account with no history in your industry, before you trust what you see.

Finding 4: Perplexity's Scarborough problem, a partial fix

In July, Perplexity resolved "Scarborough" and "Eglinton" to their namesakes in the United Kingdom on at least two of our queries, returning UK businesses for Toronto searches with no Toronto content at all. This month, on the identical Eglinton dentist query, the underlying map data still pulled up the same UK location cards. But the written answer changed: it opened with "if you mean Eglinton in Toronto" and named four real Toronto dentists before mentioning the UK alternative as an aside at the end.

That is a real improvement in what a Toronto user actually reads, even though the underlying geo-resolution bug has not been fixed. The practical advice from July stands regardless: make your location unmistakable, Scarborough, Ontario or the specific Toronto intersection, in plain text everywhere the assistant reads you.

Finding 5: ChatGPT ads pulled back hard across almost every industry

We rechecked all 50 frozen local questions logged out for a labelled ChatGPT ad, the same spot check we ran in July. July caught an ad on 2 of 50, a mismatched Ottawa injury-law ad and a windows-and-doors ad on a bathroom-renovation search. This month, zero of 50 carried an ad, including on those same two queries.

We did not stop there. Because a 50-question hyper-local pass has always drawn the fewest advertisers of any slice we measure, we also re-ran the fuller 90-question, 18-industry study behind our standalone Toronto ChatGPT Ads Index, the same five buyer questions per industry we used for June's baseline. That is where the real story is. The ad rate fell from June's 48% to about 8% this month, 7 of 90 questions. Every industry that was Saturated or High in June fell at least one tier. Finance, travel, real estate, professional and legal, and HVAC went from the top tiers straight to zero ads. Only SaaS, down from Saturated to High at 3 of 5, and online education, down from High to Medium at 2 of 5, still carry more than a stray ad, and even e-commerce and local or home services held on to just a single ad each. New advertisers we caught this round: Freshworks on a help-desk software question, The Knowledge Academy and a backlink tool called PressWhizz on two education questions, the activewear brand tentree on an e-commerce question, and Jotform again on a furnace-replacement question, the same forms tool June caught misfiring against a renovation search. Two more SaaS ads showed an advertiser we could not capture before a re-check lost them, so we logged them as ads without naming a brand rather than guess.

We are not calling this a permanent retreat off one reading. The same session-to-session volatility that flipped July's light spot check can flip a five-question sweep too, and we will not know until next month whether this is a real pullback or a quiet month. But it is the first month-over-month signal on the ads side that is bigger than noise. If you are weighing paid placement, the practical read has not changed: local-service questions remain the thinnest inventory on the whole map, and the full industry-by-industry picture, updated with this month's numbers, lives at stratezik.com/blog/toronto-chatgpt-ads-index.

Finding 6: The category leaders are consistent, established names

As in July, when multiple engines converged on the same business, it was almost always a category's established name. Gondaliya CPA was named 8 times across our 20 accounting answers, more than any other firm. Addison Pest Control led pest control with 7 mentions. Neinstein Personal Injury Lawyers, July's leader in injury law, kept the top spot with 5 mentions. In general contracting, Easy Renovation, Lanescape Architecture & Construction, and Laneway Home Builders all appeared 4 times each, all three carried over from July's leaderboard. Consistency across months, not just across engines, looks like a real signal of which businesses have built a review and citation footprint deep enough for AI assistants to keep finding them.

What this means for Toronto business owners

Six things follow from edition two.

First, the direction is good news across the board. Every engine improved, and two categories reached 100% coverage. If your business was not showing up in July, this month gave every engine more reason to find it.

Second, do not stop watching Perplexity just because it improved. An 18-point jump in one month is a big swing on a small sample, and we need at least one more reading before calling it a trend rather than a good month.

Third, if a cost-shaped or research-sounding question in your category returned nobody in July, check it again. The contracting swing shows those questions can flip to naming real businesses without warning, which means the businesses that get named early may have an advantage that compounds.

Fourth, the Claude personalization problem is now a pattern, not an edge case. If you use any AI assistant signed into a business account to check your own visibility, assume the answer may be shaped by what it already knows about you, and verify from a clean, logged-out angle whenever the stakes are real.

Fifth, keep making your location unmistakable in plain text. Perplexity's partial fix on Eglinton shows the underlying geo bug has not gone away, it has just gotten better at recovering from it in the written answer.

Sixth, ChatGPT ads remain a thin, unstable channel for hyper-local questions specifically. If you are weighing paid placement, the fuller industry map in our Ads Index is the better starting point than this tracker's 50-question spot check.

Methodology and limitations

Same 50 frozen questions as July, across the same 10 categories, run through Google AI Mode, Perplexity, ChatGPT, and Claude on August 8, 2026, five days later than the template's first-business-day target. Three answers were excluded from the percentages: two Perplexity queries with a persistent site error, and one Claude query that answered as a competitive check for our own client rather than the consumer question asked (Finding 3). Google AI Mode's full pass, five of Perplexity's queries, and ChatGPT's full pass ran in a real, signed-in browser due to a session-management slip on our end rather than the isolated, logged-out sessions the methodology calls for. We checked those answers for signs of account contamination and found none, unlike the Claude finding above, but we are disclosing the slip and tightening our process.

The ChatGPT ad-presence figure is a spot check of the same 50 local questions, not the fuller 18-industry study that backs our standalone Ads Index, which was not re-run this edition.

This is a snapshot, and AI answers vary between runs, by user, and over time. Two months of data is a trend line starting to form, not a settled pattern. The full question set and dataset are available on request.

About Stratezik

Stratezik is a Toronto marketing agency that runs on its own AI agent system. We help local businesses and startups get found and cited by AI search. Want your category or your business tracked, or a custom query run? Email dave@stratezik.com.

Sources

1. Stratezik Toronto AI Citation Tracker, August 2026: 50 frozen high-intent GTA buyer questions across 10 categories, run through ChatGPT, Perplexity, Google AI Mode, and Claude on August 8, 2026. Dataset available on request.
2. Stratezik Toronto AI Citation Tracker, July 2026 (baseline edition): stratezik.com/blog/toronto-ai-citation-tracker-july-2026.
3. Stratezik ChatGPT Ad-Presence Dataset, August 2026: the same 50 questions checked logged out for a labelled ChatGPT ad placement on August 8, 2026. Dataset available on request.
4. Stratezik Toronto ChatGPT Ads Index (June 2026 baseline plus July spot check, not updated this edition): stratezik.com/blog/toronto-chatgpt-ads-index.
5. Stratezik AEO Readiness Checker and 20-point methodology: stratezik.com/aeo-checker.