Ask ChatGPT to recommend a roofing company in Winnipeg. Ask Gemini which Winnipeg accounting firms work with trucking companies. Ask Perplexity who does commercial HVAC in the north end.
You'll get an answer. Three or four businesses, named, with reasons attached.
If your business isn’t one of them, you didn’t lose a ranking. You were never in the running, and there’s no page two to be on.
Generative engine optimization is the practice of getting your business accurately represented and cited in answers generated by AI systems: ChatGPT, Google's AI Overviews and AI Mode, Perplexity, Gemini, Claude, and Microsoft Copilot.
The term came out of a 2023 research paper from Princeton, Georgia Tech, the Allen Institute for AI, and IIT Delhi. The researchers found that specific content changes could increase a source's visibility in AI-generated answers by up to 40%. The techniques that performed best were adding relevant quotations, including statistics, and citing sources.
Note what's absent from that list: keyword density, meta tags, and every other lever the SEO industry spent twenty years pulling.
Traditional search gives you ten links and lets the user decide. A generative engine reads dozens of sources, reconciles what they say, and returns one answer naming a handful of businesses. The job changed from ranking to being chosen.
Here's the mechanic that matters, and it's the one most Winnipeg businesses are on the wrong side of.
When an AI system answers a local question, it doesn't just read your website. According to research published by AthenaHQ covering seven AI models between December 2025 and March 2026, a brand's own website appeared in roughly 16% of responses about that brand. In the other 84%, the company's own site wasn't cited at all. Reddit accounted for around 22% of citations. YouTube accounted for about 10%.
The models cited an average of roughly 12 different domains per answer, though it varies widely by platform. ChatGPT averaged close to 19 domains per response. Grok approached 27.
So the AI reads your site, your Google Business Profile, directories, review platforms, local news, and forum threads. Then it cross-checks. If your address on Yelp doesn't match your website, if your service list on a directory contradicts your homepage, if there's nothing outside your own marketing confirming you're any good, the system doesn't flag the contradiction or ask a follow-up question.
It quietly leaves you out and recommends someone it can describe confidently.
That's the Winnipeg problem. This is a mid-sized market. Most local businesses here have a thin third-party footprint: a handful of Google reviews, an outdated Yellow Pages listing, maybe a Chamber of Commerce mention from 2019, and nothing on Reddit at all. There's very little for an AI system to corroborate. Meanwhile, a Toronto competitor in the same vertical has hundreds of reviews, trade publication coverage, and active community discussion.
When someone asks an AI which Winnipeg company to hire and the local options are all uncorroborated, the model reaches for whoever it can verify. Sometimes that's a national chain with a Winnipeg location. Sometimes it's your one competitor who happens to have a strong review profile.
The good news is that thin markets cut both ways. The bar for becoming the most verifiable business in your Winnipeg category is genuinely low right now, and almost nobody local is working on it yet.
There's a lot of noise in this space, so here's our honest position on the tactic everyone is currently selling.
llms.txt is a proposed file you put at your site root telling AI systems what to read. Agencies are selling it hard. The evidence for it is weak.
A study of 300,000 domains found roughly 10% had implemented the file. Among the 50 most-cited domains in AI answers, exactly one had it. One monitoring firm reviewed over 500 million AI bot events across a 90-day window and found only 408 requests that targeted llms.txt directly. Google's Gary Illyes stated publicly in 2025 that Google doesn't support it and isn't planning to, and John Mueller compared it to the long-discredited keywords meta tag. As of 2026, no major AI provider has committed to using it as a ranking or citation signal.
We'll still add one to your site, because it costs almost nothing and may matter later as AI agents start transacting rather than just answering. But if an agency is quoting you four figures a month with llms.txt as the centrepiece of the strategy, you're paying for a file that current evidence suggests almost nothing reads.
Also worth saying plainly: Google published documentation in 2026 taking the position that optimizing for generative AI features is still, fundamentally, SEO. We largely agree. GEO is not a replacement for search optimization and shouldn't be sold as one. It's an extension of it, with a different measurement model and a much heavier emphasis on off-site corroboration.
If you’re not doing basic SEO yet, start there. We’ll tell you that instead of selling you this.
Your business name, address, phone number, hours, and service list need to be identical everywhere they appear: your site, Google Business Profile, Apple Business Connect, Bing Places, industry directories, social profiles. AI systems build a picture of your business as an entity and drop entities they can't resolve cleanly. This is unglamorous cleanup work and it is the highest-return activity in GEO.
JSON-LD schema for LocalBusiness, Service, FAQPage, Product, and Organization. This is how you state machine-readable facts about your business rather than hoping a model infers them correctly from your marketing copy.
AI systems extract passages that directly answer a question. Content that buries the answer three paragraphs into a narrative doesn't get pulled. We restructure key pages so the direct answer comes first and the supporting detail follows.
Models cite sources that say something the other twelve sources don't. Pricing ranges, process detail, local specifics, your own numbers. Generic content gets synthesized away; distinct content gets attributed.
Given that most citations come from sites you don't own, this is where the leverage is. Review volume and recency, directory presence, local publication mentions, industry association listings, and genuine participation in the community discussions your customers actually read.
We'll be direct about Reddit: it's heavily cited by AI systems and it is a terrible place to run marketing. Astroturfing gets detected, downvoted, and occasionally screenshotted. What works is being genuinely useful in threads where your expertise is relevant, and earning mentions rather than posting them.
Covered below, because it's the one that surprises people.
Start with the failure we see most often: your site may be blocking AI crawlers right now without anyone deciding to.
Cloudflare changed its default configuration to block AI bots. Plenty of sites inherited that setting. Various security plugins and WAF rules do the same. The result is that GPTBot, ClaudeBot, PerplexityBot, and Google-Extended request your pages, get refused, and the AI answers the question using your competitors instead.
Nobody notices, because it doesn't show up in any marketing dashboard.
Finding it requires reading raw server logs and filtering by crawler user-agent. Fixing it requires touching Cloudflare rules, robots.txt, and sometimes edge configuration. This is engineering work, and it's the first thing we check.
ITS Inc. is a software company that does marketing, not a marketing company that outsources development. What that means for GEO:
This is the weakest part of the GEO industry and we'd rather say so than pretend otherwise. There's no Search Console for AI answers. Anyone showing you a precise AI ranking number is estimating.
What we can measure honestly:
We run your category's prompts across the major platforms and record where you stand today: named, mentioned inaccurately, or absent. We audit crawler access, structured data, entity consistency across every listing we can find, and third-party citation footprint.
Crawler access, NAP inconsistencies, schema, technical blockers. This phase usually produces the largest single change, because most businesses have at least one hard blocker sitting in the way.
Directory cleanup, review generation systems, local PR, content that's specific enough to be worth citing.
Monthly re-runs of the same prompt set so you can see movement, plus crawler activity and referral data.
Expect three to six months before the picture changes meaningfully. Models update their training and their retrieval behaviour on their own schedule and nobody outside those companies controls the timeline.
Generative engine optimization (GEO) is the practice of getting a business accurately represented and cited in answers produced by AI systems like ChatGPT, Perplexity, Gemini, Copilot, and Google's AI Overviews. It focuses on entity consistency, structured data, answer-first content, and third-party corroboration rather than traditional keyword rankings.
Partly. Google's own 2026 documentation argues that optimizing for generative AI features is still SEO, and the fundamentals overlap heavily. The meaningful differences are that GEO puts much more weight on sources you don't own, measures citation rather than ranking, and cares about machine readability in ways classic SEO didn't. We run them together because separating them produces worse results and a bigger invoice.
Not in the way you pay for Google Ads, though AI platforms have begun introducing advertising and that will keep evolving. Organic citation is currently earned through verifiable reputation rather than bid.
Three to six months for meaningful change. Technical blockers like crawler access can produce results faster because they're binary: the crawler either reaches your site or it doesn't.
Depends on whether your customers use AI to find suppliers. For most consumer trades in Winnipeg, that's still a minority of buyers, though a growing one. For B2B, professional services, and anything with a research-heavy purchase, it's already happening. The honest answer is that the foundational work (entity consistency, structured data, reviews) improves your regular local SEO too, so it's rarely wasted regardless.
No, and anyone who guarantees it is guessing. We control the inputs. The models decide the output, they change without notice, and they don't publish their criteria.
You need to be mentioned on Reddit, which isn't the same as posting on Reddit. Earned mentions and genuinely helpful participation work. Fake accounts talking up your business get detected and can do real damage.
Serving Winnipeg and communities across Manitoba — see all our service areas.
We’ll run your category’s questions across the major AI platforms and send you the actual answers, including who’s being recommended instead of you and why. We’ll also check whether your site is accidentally blocking AI crawlers, which takes us about an hour and fixes a surprising number of problems on its own.
No charge, and no obligation to hire us afterward.
Serving Winnipeg and Manitoba.
Thank you for your interest in Impeccable Technology Solutions Inc.
We provide digital transformation, technology & development, security & compliance, professional, and AI solutions tailored to your organization.
Call or text +1 (431) 304-1000 to connect with our team.