Getting Found When Your Customer Asks an AI

What GEO actually changes versus SEO

GEO doesn't replace SEO — it's a layer on top. Most AI assistants and Google's AI Overviews source from the same organic index that classic search has always used, so rankings, backlinks and technical health still decide whether you get pulled into an answer at all. In our experience with Kando clients, the sites that show up in AI answers are almost always ones that were already ranking well organically. Skip SEO to chase GEO and you tend to end up with neither.

GEO is a layer on top of SEO — it adds to the foundation, it doesn't replace it.

In practice GEO shifts three things. Content shape moves from broad, keyword-heavy pages to narrow, direct-answer pages — one intent, one clear answer, up front. Technical exposure means adding structured data and llms.txt and making sure crawlers like GPTBot and PerplexityBot aren't blocked. And measurement shifts from keyword rankings and click-through rate toward tracking whether your brand actually gets mentioned inside AI answers. Nothing about backlinks, site speed or topical depth stops mattering; GEO just adds a second output format to optimise for. If you want the full definition, see Generative Engine Optimization, and to see where your own site stands across both layers, the Growth Audit Grader scores SEO and AI-visibility together.

Getting recommended and cited by AI assistants

Getting cited starts with making your content machine-legible and answer-shaped: clear direct answers near the top of the page, clean HTML rather than JavaScript-only rendering, structured data, and a crawlable site that welcomes bots like GPTBot and ClaudeBot. AI assistants pull from indexed, well-structured sources — Google, Bing, and their own crawls — so if a page can't be parsed, it can't be cited, no matter how good the writing is.

Beyond the mechanics you need topical depth: several pages answering related questions in your niche, not one generic overview. No vendor publishes an exact ranking formula, but the pattern that keeps showing up is that assistants lean on sources already well-indexed, structurally clear, and topically authoritative on the specific question asked. Treat it as SEO's stricter cousin — rank well organically, answer one question per page in plain language, and back claims with citable data. There's no confirmed way to game it, only ways to remove friction between your content and the model.

How you get cited — the three signals, roughly in the order a model checks them.

Kando's own GEO Technical Validator checks the same structural signals — llms.txt, schema, semantics — that determine whether a site is legible to answer engines, so you can see where the friction is before you write another word.

AI Overviews and the click-through shift

Early data and widely reported industry observations suggest AI Overviews are reducing click-through on queries where they appear, since many users get their answer without scrolling to a result. The effect appears larger on informational queries than on transactional or brand-name searches, where users still tend to click through to compare or purchase.

The response isn't panic — it's making sure you're one of the sources feeding the Overview itself.

From there, lean harder on channels less exposed to this shift — email, communities, and direct brand search. Treat AI-answer citations as a new visibility channel worth earning even without a click, while keeping SEO fundamentals strong for the traffic that still converts through a visit. This is diversification, not a reason to abandon what already works.

Why JSON-LD and schema matter

JSON-LD structured data labels your content explicitly — this is a question, this is an answer, this is an organisation, this is a price — so machines don't have to guess meaning from layout. AI crawlers and the search indexes they draw from can parse structured data faster and more reliably than freeform HTML, which lowers the odds your content gets misread or skipped. It won't force a citation on its own, but it removes a common failure mode: a good answer a model simply can't parse cleanly.

For a service business, the highest-value types are Organization (or LocalBusiness/ProfessionalService for location-based work), FAQPage for Q&A content, and Service for each offering, with Article or WebPage markup on blog and answer content. Priority beats completeness — a well-tagged Organization plus FAQPage combination on your core pages typically does more than exhaustive markup applied inconsistently. Broken schema can be worse than none, so validate before you publish. Kando's Schema Generator & Validator builds and checks this markup without needing a developer.

What llms.txt does, and whether it helps

llms.txt is a plain-text file at your site's root that gives AI crawlers and language models a curated map of your most important pages, written in plain language rather than HTML. It works like a robots.txt for meaning instead of permissions — pointing models toward the content you'd most want summarised or cited.

Here's the honest part: whether it moves the needle is still genuinely unproven, and no major AI vendor has confirmed it directly influences citation. It's cheap to ship and can't hurt, which is why Kando publishes one on kandomarketing.com alongside JSON-LD schema. Treat it as good hygiene, not a silver bullet — and be suspicious of anyone selling it as a guaranteed shortcut to being cited.

Should you block or allow AI crawlers

For most businesses that want visibility in AI answers, the trade-off favours allowing crawlers like GPTBot, ClaudeBot and PerplexityBot in robots.txt — blocking them guarantees you can't be cited, since these models mostly answer from what they, or their search partners, have crawled and indexed. The one exception is publishers whose entire business is the content itself, where training-data and traffic-cannibalisation concerns can outweigh the visibility upside.

Kando's own site welcomes AI crawlers deliberately, on the logic that being unreachable is the one guaranteed way to lose every AI-search citation. Watch for the quiet own-goal here: an over-broad robots.txt rule that blocks these bots by accident is one of the most common ways businesses lock themselves out without realising it.

Digital PR and mentions as GEO fuel

AI assistants tend to trust and repeat claims that show up consistently across multiple independent, credible sources, so a mention in a trade publication, a stat picked up by other sites, or a listing on a respected industry roundup all function as corroboration a model can cross-check. This is the same mechanism that made backlinks valuable for SEO, just applied to trust rather than link equity.

A single self-published claim on your own site is weaker evidence to a model than that same claim echoed by three unrelated outlets — which is exactly why digital PR remains a core, not optional, part of a GEO strategy. Original data amplifies this: a proprietary survey or a benchmark study becomes the single source for a fact, so any assistant answering a related question has to point back to you. Even a survey of fifty clients can out-earn a much longer generic article for citation purposes, because uniqueness beats volume in an AI-answer context.

Keeping content fresh for AI

There's no fixed universal cadence, but the practical rule is to revisit high-value pages often enough that facts, prices and examples never go stale, and to treat anything tied to fast-moving topics — pricing, tools, statistics — on a shorter cycle than evergreen explainer content. Quarterly reviews of top-traffic pages is a reasonable default for most small teams; monthly for genuinely fast-moving categories.

Freshness signals seem to matter more for AI answer engines than they historically did for classic SEO rankings, since a model surfacing outdated pricing or a dead statistic is a visible, embarrassing failure. Kando's Content Decay Detector flags pages that are quietly losing traffic, which is often the first sign content has gone stale.

Do FAQ and Q&A pages still work

They still work — arguably better than ever, because FAQ and Q&A pages are structurally exactly what AI answer engines want: one clear question, one direct, self-contained answer, easy to lift and quote. The key is discipline: one distinct intent per page or entry, not a long list of loosely related questions crammed onto a single URL.

This is the model Kando's own Answers wiki follows — each question gets its own page with FAQPage schema attached, so it can be indexed and cited on its own terms rather than buried in a longer article. Open with the answer, not the setup: the first sentence after a question should state it in full, self-contained language that still makes sense pulled out of context. That's the sentence a model quotes. You can browse the Kando Answers wiki to see the pattern applied end to end.

Building topical authority

Building topical authority in practice means covering a subject in depth and breadth before moving to the next one: map out every real question a buyer or researcher asks about the topic, publish direct answers to each, and interlink them tightly so both search engines and AI crawlers can see the cluster as a coherent whole. One deep, interlinked cluster of twenty pages typically outperforms twenty unrelated pages scattered across different topics.

Consistency over time matters as much as volume — a cluster that keeps getting updated signals ongoing expertise, not a one-time content push. The signals that seem to carry weight are the concrete, verifiable ones: named authors with real credentials, author and organisation schema, consistent facts across your site and third-party mentions, and original data or first-hand experience. Because AI assistants often cross-check claims against multiple sources before surfacing them, inconsistency quietly hurts you more in the AI era than it did in classic SEO.

How to measure AI-search visibility

Measurement here is still emerging, and nobody has a clean analytics equivalent of Google Search Console yet. The practical approach right now is manual and repeated: ask ChatGPT, Perplexity, Gemini and Claude your target questions on a schedule, log whether and how your brand gets mentioned, and track it as a directional trend — sometimes called tracking share of model — rather than a precise number.

A few emerging tools attempt to automate this at scale, but in our experience the manual-prompt method is still the most reliable starting point for a small or mid-size business. It's worth knowing the engines differ under the hood: Perplexity is the most search-like and rewards fresh indexing, Gemini leans heavily on Google's own index, and ChatGPT's retrieval behaviour mixes Bing-indexed content with its own training data. Strong, well-structured, well-cited SEO is the common denominator across all of them.

How to audit your site's AI-readiness

Diagnose before you fix. Start with the technical layer: check whether AI crawlers can actually read your pages — server-side rendering, no accidental robots.txt blocks — whether structured data and llms.txt are in place, and whether your core pages open with direct, quotable answers rather than marketing preamble. Server-side rendering matters because many AI crawlers don't fully execute JavaScript; if your content only appears after client-side rendering, a crawler may see a near-empty page and have nothing to cite. Kando ships kandomarketing.com server-side rendered for exactly this reason, alongside llms.txt and JSON-LD.

Then check the content layer: is there topical depth, consistent facts, and credible third-party corroboration for your key claims? Kando built two free tools for exactly this — the GEO Technical Validator checks structured data, llms.txt and semantics, and the Growth Audit Grader gives a broader SEO and AI-visibility score across a full site.

The biggest GEO mistakes to avoid

The most common mistake is treating GEO as a separate initiative from SEO instead of a layer on top of it — chasing llms.txt and schema while ignoring weak organic rankings, which are still the foundation AI answers draw from. Close behind: blocking AI crawlers by accident in robots.txt, burying direct answers under marketing copy instead of leading with them, and JavaScript-only rendering that leaves crawlers with nothing to read.

The quieter mistake is inconsistency — the same claim, told two different ways on two different pages.

Publishing a stat or claim on one page that contradicts another page or a third-party mention undermines exactly the cross-checking AI assistants increasingly do before citing a source. Fix the foundation, lead with the answer, keep your facts consistent, and let the model do the rest. If you only do one thing after this guide, run your site through the GEO Technical Validator and fix whatever it flags first.