Search still sends business to your website. It just sends less of it, and it sends it differently. SEO and GEO now have to run as one program rather than two disciplines with separate budgets, because the page that ranks well in Google is usually the same page an AI assistant reads before it writes its answer. Any seo company still selling positions alone is going to start hearing uncomfortable questions from clients who watched their traffic slide while their rankings held steady.
Generative engine optimization is the practice of structuring content so that ChatGPT, Google AI Overviews, Perplexity, Claude, and Copilot cite it in their responses. Traditional SEO earns you a position on a results page. GEO earns you a mention in the answer, which is often where the decision is actually made now.
GEO is not a replacement for SEO, and treating it as one is expensive
There’s a version of this conversation that goes “SEO is dead, pivot to AI.” It’s wrong, and it costs clients money.
Every serious analysis of how generative engines select sources points back to the same signals SEO people have worked on for years. Strong organic positions, clean crawlability, credible backlink profiles and clear topical authority remain the raw inputs. An AI model does not discover a page that Google cannot index. What GEO adds is a second layer of requirements on top: content that can be extracted as a direct answer, claims that are specific enough to be worth quoting, and a brand identity consistent enough across the web that a model recognizes it as an entity rather than a string of words.
So the practical shift is additive, not substitutional. You keep doing technical audits and link acquisition. You add answer-shaped content, structured data at scale, source transparency, and regular testing of how your client’s brand appears inside AI outputs.
What marketers are actually spending on
Forty-four percent of surveyed marketers now name AI-powered search as their primary source for digital insights and trends, ahead of the 31% who still rely on conventional search. That single flip is the clearest signal of where research behaviour has moved.
A June 2026 survey of 250 digital marketing professionals found that 89% intend to increase investment in AI-friendly content formats and monitoring. Seventy-eight percent expect generative optimization to matter more than traditional SEO by 2028. Sixty-five percent have already started reworking existing content libraries for AI summaries.
The number that should get attention is the fourth one: 52% report measurable declines in organic traffic they attribute to AI overviews answering queries without a click.
That last figure is the one worth taking to clients, because it reframes GEO from a speculative add-on into a defensive necessity. Half the market is already losing sessions.
Within the emerging service categories, AI visibility analytics leads at roughly a third of the market, followed by content optimization built for large language models at just over a quarter, with brand citation monitoring, AI answer placement and conversational search work splitting the rest. That distribution indicates where agencies believe the billable work lies.
Where the outside research agrees, and where it complicates the story
Gartner forecast a 25% contraction in traditional search engine volume by 2026, a projection published in early 2024. We’re now inside that window, so the honest move is to check it against what actually happened in your clients’ analytics rather than repeating it as a prediction. Semrush has tracked AI Overviews appearing in roughly one in six searches, though that share has moved considerably and any figure you quote should carry a date.
Here’s the part most agency content skips. Gartner’s consumer research has also found substantial distrust of AI-generated results, with a majority of respondents expressing doubt about accuracy, and a meaningful minority reporting that AI summaries lengthen their research rather than shortening it. Only about a third of American consumers rated generative tools as equivalent to a conventional search engine.
Read those together, and the picture is less apocalyptic than the traffic charts suggest. People are using AI to orient themselves and then verifying elsewhere. That verification step is your opportunity. If the AI cites your client and the click goes to a competitor’s better page, you’ve won the citation and lost the sale.
BrightEdge’s research on this transition found that SEO teams are leading AI initiatives in the majority of organizations studied, which aligns with what we see in practice. The people who already understand crawl budgets and schema are the people best positioned to do this work.
Four layers of work that hold up under scrutiny
Content. Write for extraction. That means real question-and-answer structure, definitions stated plainly in the first sentence of a section, original data where you have it, named authors with verifiable credentials, and a traceable source behind every claim. Fact density matters more than word count. A 900-word page with eight specific, checkable statements will get cited more often than a 3,000-word page of generalities.
Technical. Schema markup deployed properly across page types, not just on the homepage. Core Web Vitals in the green, which you can measure against Google’s published thresholds. Clean indexation, fast loads, HTTPS, mobile-first rendering. The schema.org vocabulary is still the reference for markup types, and getting Organization, LocalBusiness, FAQPage and Article right does more for AI comprehension than most content tweaks.
Authority. Consistent brand mentions across your site, social profiles, video platforms, review sites and industry publications. Generative models place heavy weight on repetition and consistency when deciding whether a brand is a credible source on a subject. Editorial backlinks and earned media still do the heavy lifting here.
Testing. Query the major AI tools directly with the questions your clients’ customers ask. Record which sources get cited. Do it monthly. This is tedious, and it’s the only reliable feedback loop that currently exists, since none of the generative platforms report referral data the way Search Console does.
Anthropic’s and Google’s own guidance on helpful content, including Google’s documentation on creating people-first content, remains a reasonable baseline for what these systems reward.
The Toronto and GTA angle
Local businesses face a sharper version of this problem. When someone asks an AI assistant for the best plumber in Mississauga or a family lawyer in North York, the model assembles its answer from Google Business Profile data, review content, directory citations and location pages. Inconsistent NAP data across those sources no longer just weakens local pack rankings. It gives the model conflicting information and makes it less likely to name the business at all.
For clients across the GTA, the highest-value work is usually unglamorous: reconciling citations, properly filling out Business Profile categories and service lists, generating reviews that mention specific services by name, and building location pages with genuine local detail rather than the same paragraph with the city swapped in. Our Google Ads work often surfaces the exact question phrasing that local customers use, and feeding those queries back into content is one of the cheapest wins available.
Sequencing the work
For agencies rolling this out across a client base, the order matters more than the tactics:
- Audit current organic standing and, separately, baseline how often each client’s brand appears in AI answers for their top 20 commercial queries.
- Prioritize pages by commercial value, then rework the highest-value ones for extraction before touching anything else.
- Apply technical fixes, schema first, since it has the broadest effect for the least effort.
- Launch authority work: publishing cadence, editorial links, earned mentions.
- Build reporting that shows both traditional and generative metrics side by side, so clients see the whole picture.
- Train the team, then repackage services and pricing to reflect the added scope.
- Review quarterly and adjust, because the platforms change faster than annual planning cycles allow.
The problems nobody has solved yet
GEO attribution is genuinely unsolved. There’s no equivalent of Search Console for AI citations, so proving return on investment relies on proxy metrics and correlation. Be honest with clients about that rather than manufacturing certainty.
Consumer distrust cuts both ways. If a model misrepresents a client’s services or pricing, there’s currently no formal mechanism for correction. The mitigation is making the correct information so clear, so structured, and so consistently repeated across the web that the model has little room to invent an alternative.
And resourcing is real. Most agency teams do not have spare capacity for monthly AI visibility testing on top of existing deliverables. Start with three pilot clients, prove the process, then scale.
The agencies that come out of this period ahead will be the ones treating AI visibility as a measurable deliverable with a defined process behind it, not a bullet point added to an existing proposal. If you want to know where a specific site currently stands, a baseline audit covering both organic position and AI citation rates will tell you in about a week whether you’re building from strength or playing catch-up. That’s the conversation worth having now, before the gap widens. Book a free audit with our team, and we’ll show you exactly where your visibility stands.
