The State of Lodge Websites in the AI Search Era

You built the lodge on word of mouth. A guest had the week of their life, told a buddy at work, and that buddy called in February. That worked for thirty years, and it still works — just for fewer people each season, and for guests who are getting older right alongside you.

The website was never the point. It’s the job that waits until the season’s over, and by the time the season’s over you’re tired. So it sits there: a few pages, the rates, a gallery, a phone number. It’s not broken. It loads. People who already know your name can find you.

The trouble is that fewer of the people deciding where to fish this year start by knowing your name.

They start by asking. And more and more, they’re not asking Google the old way — they’re asking a tool that answers back in a paragraph. “Where’s good walleye fishing in Northwest Ontario for a group of six in late June?”

The tool reads the web, picks a few lodges it can describe clearly, and names them. The person planning the trip reads that paragraph and makes a shortlist before they’ve clicked anything.

So the question worth sitting with isn’t whether your site looks nice.

It’s this: when a tool like that goes looking for a lodge to recommend for a real trip, is there anything on your website it can lift out and quote with confidence — or does it move on to the lodge whose page gave it a cleaner answer?

We went looking for the pattern.

What we looked at

We ran a surface crawl of 41 fishing-lodge, outfitter, and camp websites across Northwest Ontario — the kind of drive-to and fly-in operations that compete for the same guests you do.

The crawl measures a few plain things about a website’s text: how much of it is unique versus repeated across the site’s own pages, how many pages are too short to answer anything in depth, how many pages go deep enough to fully answer a question, and — before any of that — whether an automated reader can get into the site at all.

This is a surface read, not a full audit. It doesn’t judge whether the writing is good or whether the lodge is any good — plenty of these are excellent operations.

It measures one narrow thing: how clearly each site hands a machine a distinct, complete answer it can trust and repeat. That happens to be the exact thing an AI tool needs before it will name you.

Here is what the field looks like.

The findings

(Figures below describe the Northwest Ontario sites we examined — a surface-level pattern of the field. Of the 41 sites, 34 could be read completely; the content figures describe those 34.)

➤ Every single lodge site we could fully read showed at least one of three answer-clarity gaps — pages that mostly repeat each other, pages too thin to answer anything, or almost no pages deep enough to fully answer a guest’s question. All 34 of 34.

➤ Nine out of ten sites (91%) had a third or more of their pages reading as near-duplicates of their own other pages — the same blocks of text, or the same page living at two web addresses. On the middle site, more than half the pages overlapped this way.

➤ On close to six in ten sites (59%), more than half of all pages were internal near-duplicates.

➤ Two-thirds of sites (65%) averaged under 500 words per page — often not enough room to be the full, standalone answer to a specific question a guest is actually asking.

➤ On more than half the sites (53%), fewer than one page in ten went deep enough (800+ words) to serve as a complete answer. The deep, quotable page is the exception, not the rule.

➤ Seven of the 41 sites (17%) couldn’t be fully read by an automated visitor at all. Three returned an outright “no entry” response. A site an automated reader can’t open is a site an AI tool can’t quote.

None of this is a character flaw.

It’s what happens when a website gets built once, added to in a hurry over a few seasons, and then left to run the business while the owner runs everything else. It’s the most normal thing in the world. It’s also, right now, an opening.

What the numbers are really saying

It’s tempting to read “duplicate pages” as a scolding — as if a search engine is going to dock you points for it. That’s the old story, and it’s not the useful one. Set the penalty idea aside.

Here’s the mechanism that actually matters now. When someone asks an AI tool a specific question — “which lodges have wheelchair-accessible cabins,” “where can a group of eight book a shore-lunch trip in the third week of June,” “what’s the walleye limit and season on that chain of lakes” — the tool goes looking for one page that answers that one question clearly and completely. It wants to lift a clean, confident answer and attach your name to it.

If your rates page, your fishing page, and your “at the lodge” page all say roughly the same handful of things, the tool doesn’t find one strong answer — it finds five faint echoes and can’t tell which to trust. If your pages are short, there’s no complete answer to lift. If the deep, specific answer to that guest’s question was never written, there’s nothing there to quote at all.

So the real question underneath the data isn’t “how much duplicate text do I have.” It’s quieter than that:

When someone asks the exact question your lodge is the best answer to — the one you’d answer perfectly if they called — is that answer written down anywhere on your site, once, clearly, in full? Or does it only exist in your head, on the phone, for the people who already knew to call?

For most of the field, the answer is: it lives on the phone. Which means it’s invisible to the tool doing the recommending.

The cost of staying exactly where you are

Nothing breaks if you leave it.

That’s what makes this one easy to keep putting off.

The site keeps loading, the phone still rings, the regulars still come.

But picture the guest you don’t hear from.

A group of four in their thirties — the exact younger group you’d love more of — planning a first Northwest Ontario trip. They don’t have a lodge in mind. They ask a tool. It names three lodges it could describe clearly, with pages that answered their questions in full. Your lodge answers those questions beautifully — on the phone, if they call.

They never call, because they never saw your name.

They booked one of the three. You didn’t lose that booking on the water or on price or on the quality of your operation.

You lost it because, at the one moment a machine was deciding who to name, your best answer wasn’t written down where it could be read.

Do that math across a season of shortlists you never appear on, and across the younger guests who plan this way by default, and the number gets real — you can already feel the shape of it.

The part that should feel like good news

Look back at the field. Nearly every site has the same gap. That’s not an indictment — it’s the opportunity, and it’s rare.

It means the lodge that writes a handful of clear, complete, distinct answers to the questions guests actually ask — and makes sure a machine can read them — doesn’t have to beat a field of polished competitors.

It has to beat a field where almost no one has done this yet.

In our case studies, a lodge that got this right is already being named by Google’s AI answer as the source for a real fishing-season question — a first-mover lead none of the regional operators we examined had matched. (This is verified for that lodge and query; the broader growth pattern is directional.)

➤ Being early to this is worth more than being early to most things, because of how these tools behave: once a tool learns your site is a reliable place to get a clean answer, it tends to come back and name you again. The lead compounds. The window where the whole field is wide open does not stay open forever.

What we’re calling this

The thing we’re measuring has a name, so we can talk about it plainly.

AI Citation Visibility is the degree to which a business is surfaced and named as a recommended answer by AI search tools — Google’s AI Overviews, ChatGPT, Perplexity, Claude — when someone asks a question it should be the answer to.

It’s a different game from ranking.

Ranking asks where does my blue link sit on the results page. AI Citation Visibility asks when the machine writes the answer, does it name me. You can rank respectably and still never get named. And you can get named — pulled straight into the answer a trip-planner reads first — long before you’d ever climb to the top of a crowded results page. For a lodge that’s been getting by on a site that merely exists, that second path is the faster one, and right now it’s the emptier one.

The next small step

If reading this left you wondering where your own site sits in that field, that’s the place to start — with your site, not the category.

There’s a short check for it: Would AI recommend your lodge? — a Findings Audit that asks the real buyer questions your guests would ask a tool, and shows you, in plain terms, whether your site currently gives a clean answer or gets passed over. No rebuild, no commitment. Just a clear picture of where you stand while the field is still this open — so the decision about what to do next is yours to make with the facts in front of you.

The lodges that move on this first are going to be the ones a machine has already learned to name. There’s room in that group right now. There won’t always be.