Getting cited by ChatGPT and AI search comes down to three things: the crawler behind the assistant must be allowed to fetch your pages, the page must answer a specific question in plain language near the top, and your structured data must actually render in the HTML. For Hamilton, Ontario businesses the third one fails most often, and the SEO plugin will usually report that it is fine.
What actually decides whether ChatGPT cites your website?
Three gates, in order. The crawler has to receive a 200 response, the answer has to be extractable as a short block of text, and the machine readable layer has to exist in the rendered HTML. Fail the first gate and nothing else matters, because the assistant never sees the page at all.
This is the part that surprises owners who have been doing traditional SEO properly for years. Backlinks and domain age are not what an assistant consults at the moment it writes an answer. It retrieves a small set of pages, reads them, and quotes whichever one states the answer most plainly. A page that ranks fourth but answers the question in its opening sentence gets named ahead of the page that ranks first and buries the answer beneath a company history.
So the work splits in two: make the page machine readable, then make the answer quotable. In our experience the sites that lose out on AI search in Hamilton are rarely losing on writing quality. They are losing on plumbing nobody has looked at.
Why does structured data that looks configured still show nothing to AI search?
Because of where the plugin stores it. A failure mode we check for on every audit is structured data saved in the wrong storage format, which renders into the page as a node with no type at all. The plugin interface reports the schema as present and correct, while a machine reading the page sees nothing.
It is worth sitting with how quiet that failure is. There is no error, no warning and no broken layout. The schema tab is filled in, the fields are populated, and the preview looks right. The page still carries a block of JSON-LD, so a casual glance at the source finds structured data and moves on. What it does not carry is the one property that makes the block mean anything: the type declaration that tells a parser whether it is looking at an article, a local business or a list of questions. Without it the data is just an anonymous bag of text, and every consumer of structured data, Google included, discards it.
The only way we catch this is by reading what the server actually sends rather than what the editor claims. That is a different habit from the one most agencies have, and it is the single highest value check we run on a site that is already well written. If you want to confirm your own, Google publishes the rules for valid markup in its structured data documentation, and its Rich Results Test reads the live URL instead of the plugin.
Takeaway: never trust a settings screen. The only trustworthy answer is what the rendered page contains.
How do you check what AI search really sees on your page?
To check what AI search really sees, request the page with an assistant crawler’s user agent and confirm a 200 response, search the raw HTML for a JSON-LD block, confirm every node declares a type, validate the live URL and read your first sixty words aloud. Work through these in order, because each step depends on the one before.
- Request the page the way a crawler does, not in a browser. Use a command line request with an assistant user agent and confirm you get a 200 response rather than a 403.
- Save the raw HTML the server returned and search it for the application/ld+json block. If there is no block, you have no structured data, whatever the plugin says.
- Read every node in that block and confirm each one declares a type. A node without one is invisible, and it is the most common defect we find.
- Run the live URL through a validator that fetches the page itself, so the result reflects what a parser receives.
- Read the first sixty words out loud. If they do not answer the question in the title on their own, an assistant has nothing short enough to quote.
- Repeat the whole check after any plugin update, theme change or CDN setting change. All three can silently alter what gets served.
Which signals should a Hamilton business check first?
Check structured data, crawler access and your opening answer first, then question coverage and freshness. For each signal, the place an owner usually looks, such as the schema tab or robots.txt, is not the place the decision is made, such as the rendered JSON-LD or the response code each crawler receives at the edge.
| Signal | Where owners usually check it | What actually decides it | How we verify it |
|---|---|---|---|
| Structured data | The schema tab in the SEO plugin | Whether the rendered JSON-LD contains a typed node | Read the raw HTML the server returns, then validate the live URL |
| Crawler access | robots.txt | The response code each assistant crawler receives at the edge | Request the page with each crawler user agent and record the status |
| The answer itself | Word count and keyword placement | Whether the first sixty words stand alone as an answer | Read the opening in isolation, with the rest of the page hidden |
| Question coverage | Whether the topic is mentioned somewhere | Whether a heading states the exact question being asked | Compare headings against the questions customers actually ask |
| Freshness | The published date | Substantive change to the material, not a new timestamp | Diff the content, not the date |
Can you optimise for AI search while your site blocks AI crawlers?
No. If the crawler cannot fetch the page, nothing on it can be cited. We tested the major assistant crawlers against a Cloudflare protected WordPress site and recorded a 403 Forbidden for every one of them, while Googlebot and Bingbot received 200 OK. Check crawler access before you spend anything on content.
The reason this goes unnoticed for so long is that traditional search keeps working perfectly while it happens. Rankings look normal, traffic looks normal, and the entire assistant channel is quietly answering with a competitor website. We walked through the crawler test and the settings involved in our small business website checklist for Stoney Creek, including which crawler categories to allow if you want to stay citable without handing your content to model training.
How should you write a page so an assistant can quote it?
Write the page so an assistant can lift the answer without editing: lead with two or three factual sentences, phrase headings as the question, put comparisons in real tables, be specific, answer one question per page properly and keep facts out of images, sliders and tabs. Once the plumbing is right, the writing decides whether you get named.
- Lead with the answer. Two or three factual sentences at the top, before any preamble about who you are. An assistant needs a block it can lift without editing.
- Phrase headings as the question. Not Pricing, but what does a new website cost in Hamilton. The heading is the retrieval handle.
- Put every comparison in a real table. Structured rows are extracted far more readily than the same facts described in a paragraph.
- Be specific and attributable. Named numbers, named places, named methods. Vague claims about what studies show are not quotable, because there is nothing to attribute.
- Answer one question per page properly rather than ten questions thinly. Thin coverage of a wide topic is the fastest way to be passed over for a page that goes deep.
- Keep the facts where a parser can find them. Text in an image, a slider or a tab that loads on click may never be read at all.
None of this is separate from ranking well. The same structure that gets you quoted is what earns the local SEO results Hamilton service businesses are usually after in the first place, which is why we treat it as one job rather than two budgets.
What does it cost to get a Hamilton site ready for AI search?
It depends on whether you need the checks or the rebuild. The audit work above is a few hours on most sites, and on a well built site the fixes are configuration rather than development. Where it gets larger is when the content itself has to be restructured, or when the underlying build hides its text from parsers.
Our builds start at Starter from $1,499 for a focused site, Growth from $2,999 where there is a real content structure and technical SEO scope to get right, and Commerce from $4,999 for selling online. Schema, crawler configuration and answer led page structure are part of the build in each case, not an upsell afterwards. If your site is recent and sound, you may need nothing more than the verification pass and some rewriting, which is worth saying plainly before anyone quotes you for a rebuild.
Whatever you spend, the maintenance side matters more here than it does for traditional SEO, because a plugin update can undo a schema fix without anyone noticing. We covered what tends to break and what it costs to put right in our guide to website maintenance for Hamilton businesses.
Does AI search replace Google for Hamilton businesses?
Not yet, and not evenly. Most Hamilton service businesses still get the majority of their enquiries through Google, maps and referrals. What has changed is that a growing slice of research now happens inside an assistant, and only pages that are readable, structured and specific get named in those answers.
The practical position we take with clients is that this is not a new channel to fund, it is a readability standard to meet. A page that an assistant can parse and quote is also a page Google can understand, a page a screen reader can navigate and a page a hurried customer can skim. The work compounds instead of competing.
If you want someone to run the verification pass on your own site and tell you plainly which of the gates you are failing, that is part of our technical SEO services in Hamilton, and you can book a technical review with our Hamilton team to get the actual response codes and JSON-LD from your pages rather than a guess.