Should Guides Use AI Content?

Most arguments about this are about quality, and quality is the least interesting part. The harder fact is that generated text is nobody's property, and a guide who fills a website with it has built something anybody may copy, including the operator on the next stretch of water.

That is not an opinion. It is the settled position of the office that decides what is protected, stated in a policy document with worked examples, and it changes the calculation for a small business whose website is one of its few real assets.

What the Copyright Office actually said

In March 2023 the office issued a statement of policy clarifying how it examines and registers works containing material generated by artificial intelligence technology. Its reasoning runs through one question, which is whether the traditional elements of authorship were produced by a human being.

Where a technology receives only a prompt from a human and produces complex written, visual or musical work in response, the guidance concludes that those traditional elements are determined and executed by the technology rather than the user. Its explanation of why is the sentence worth remembering: users do not exercise ultimate creative control over how such systems interpret prompts and generate material.

Then the analogy that settles it. Prompts, the office says, function more like instructions to a commissioned artist: they identify what the prompter wishes to have depicted, but the machine determines how those instructions are implemented in its output.

Everything else a guide does to be found is at the getting-booked hub, and it is worth reading this before deciding to fill any of it with generated text.

Three Rivers Ranch, ID

The worked example

The guidance uses one that transfers directly. If a user instructs a text-generating system to write a poem about copyright law in the style of Shakespeare, she can expect text that is recognisably a poem, that mentions copyright, and that resembles the style. But the technology will decide the rhyming pattern, the words in each line, and the structure of the text.

Substitute your own case. Ask for eight hundred words about tarpon fishing in June and you can expect something recognisably about tarpon, mentioning June, in the register you asked for. The machine decides the sentences, the order, the emphasis and every specific claim in it. That is the whole of what a reader is judging.

The conclusion follows: when the technology determines the expressive elements of its output, the generated material is not the product of human authorship, is not protected by copyright, and must be disclaimed in a registration application.

Nobody owns it, which is the practical problem

Take that out of the registration context and into a guiding business.

Your site is one of the few things you own outright that compounds. Pages accumulate, rank, and get read for years. A page whose text nobody authored has no owner, which means a competitor may lift it verbatim and you have no claim to make. Not a difficult claim. No claim.

That is a strange asset to build a business on, and it is the argument that persuades guides who are unmoved by every argument about voice. What you own and what you merely possess is a distinction worth being clear about generally, and it is worked through in the ownership piece.

The line is drawn case by case

The guidance is careful not to say that using these tools disqualifies a work, and that nuance matters for anybody who uses one sensibly.

The office says it will consider whether the contributions are the result of mechanical reproduction or instead of an author's own original mental conception to which the author gave visible form, and that the answer depends on the circumstances, particularly how the tool operates and how it was used. It calls this necessarily a case-by-case inquiry, and notes that many technologies are marketed as artificial intelligence without functioning the same way for these purposes, so the analysis is fact specific.

What that means in practice is that a guide who writes the substance and uses a tool to tidy it is in a different position from one who publishes what the machine produced. The first authored something. The second commissioned it and cannot register it.

The disclosure duty, and where it comes from

There is an obligation attached, and it rests on an old and broadly worded provision. Applicants have a duty to disclose the inclusion of generated content in a work submitted for registration, and to provide a brief explanation of the human author's contributions.

The hook is the registration statute, which lists what an application must contain: the claimant's name and address, the author's name and nationality, whether the work was made for hire, how a claimant who is not the author obtained ownership, the title, the year creation was completed, publication details, and for a compilation or derivative work an identification of the preexisting material plus a statement of what is newly covered.

Its final item is the catch-all the guidance relies on: any other information regarded by the Register of Copyrights as bearing upon the preparation or identification of the work, or the existence, ownership or duration of the copyright. Individuals using these tools may still claim protection for their own contributions, using the standard application and identifying what those contributions were.

How the office reached this position

Not recently, and not theoretically. The guidance describes a 2018 application for a visual work that the applicant themselves described as autonomously created by a computer algorithm running on a machine. It was refused because the examiner found the work contained no human authorship, and after administrative appeals the office's review board affirmed, on the ground that it was made without any creative contribution from a human actor.

The guidance also reaches back to a much older formulation, that works must be created by a person rather than by a machine. This is not a new rule invented for the current wave of tools; it is an old rule meeting them.

The second problem, which is worse

Ownership is the argument that changes minds. Accuracy is the one that costs money.

A guiding site makes claims of a specific kind: seasons, species, sizes, regulations, what a licence costs, whether one is needed. Every one of those is checkable, several of them change annually, and a confident wrong answer on a licensing question sends somebody onto the water without what they need.

Generated text is fluent about exactly these things and has no way to know which are true for your water this year. A machine writing about your river will happily assert a season that closed two years ago, in a sentence better constructed than anything you would have written, which is precisely what makes it dangerous.

That is also the reason this trade is a poor candidate for volume content, and why the standards that apply to a fishing report apply doubly here, as set out in the fishing reports piece.

What it is genuinely useful for

Structure, before you write. Asking what somebody planning a first trip would want answered produces a decent list, and the list is not the writing.

Editing your own draft. Tightening, cutting, catching the sentence you wrote twice. You authored it; the tool tidied it, which the guidance's own framing treats very differently from generation.

The mechanical parts nobody reads for pleasure. Page titles, descriptions, image alternative text, and the repetitive elements that have to exist and carry no judgement.

And as a reader. Pasting in your own page and asking what question it fails to answer is a cheap review, and the answer is often the thing that was missing.

What it should never write

Anything about your water. The specific detail is the only thing on your site a competitor cannot copy, and it is the one thing a machine cannot supply because it has never been there.

Anything about regulations, licences, seasons or limits. Those belong to an authority, they change, and the correct move is a link to the authority rather than a paragraph of your own.

Your own story. The paragraph explaining who you are and why you guide is the highest-converting text on most guide sites, and a generated version reads as though written by somebody who has never met you, because it was.

And anything presented as a client's words. A generated testimonial is a different and more serious problem than a generated paragraph, and it is treated properly in the testimonials piece.

Lodge & Ranch at Chama, NM

The dictation route

The fastest way to produce a lot of writing that is genuinely yours has nothing to do with any of this, and almost nobody in the trade uses it.

Talk. Five minutes on a phone recorder about what happened on Tuesday, what the water did, why you moved, what the client found difficult. Transcribe it, cut the false starts, put in the paragraph breaks. What comes out is specific, has your rhythm, contains details a machine could not invent, and took less time than writing three careful sentences.

A tool can do the cleaning up without touching the substance, which is precisely the position the guidance treats differently from generation: you determined what the piece says and how it says it, and something else fixed the punctuation.

Guides who do this for a season end up with more usable material than any content plan produces, because the raw supply is a thing they were doing anyway. The trip photos work the same way, and pairing the two is what turns a season into a website, as described in the trip photos piece.

Volume is the wrong target

The pitch for generated content is always about quantity, and quantity is the wrong objective for a business that serves one stretch of water.

A guide is not competing with a publisher. The pages that matter are few and specific: the waters you actually fish, the trips you actually run, the questions you are actually asked. Five of those, written properly and kept current, outperform fifty about places you have never launched a boat.

There is a second cost that is easy to miss. Pages accumulate maintenance. Anything asserting a season, a price or a rule has to be revisited, and a guide with fifty generated pages has fifty liabilities they did not write and cannot remember the contents of. The volume that felt free in January is the reason nothing gets updated in March.

And thin pages about waters you do not know attract readers you cannot serve, which produces enquiries you have to decline. That is worse than no traffic, because declining takes time and disappoints somebody.

A test before publishing anything

Read the page and ask a single question: could the guide on the next river have published this word for word?

If yes, it says nothing about you, it is not protected, and it will not make anybody choose you. If no, because it names a launch, a month, a fish that behaves a particular way, or a mistake you watched somebody make, then it is doing the job and it belongs to you.

Run the same test on the paragraph about who you are, which is the one guides most often outsource and the one that most needs to be theirs.

Then check every checkable claim on the page against the authority that issues it, and link out rather than restating. Those two habits, the specificity test and the outward link, are most of what separates a guiding site that earns trust from one that reads like every other result on the page.

What experienced guides do differently

They write the parts only they can write and leave the rest short. A page with four honest paragraphs beats one with twelve generated ones, and it is faster to produce.

They dictate rather than type. Talking through a day on the water for five minutes and cleaning up the transcript produces text that is unmistakably theirs, and it is the fastest route to volume that does not sacrifice ownership.

They keep a file of their own sentences, the descriptions and explanations that have worked, and reuse them deliberately across pages, emails and replies.

And they check every checkable claim against the authority rather than against another website, which is the discipline that separates a site that ranks from one that quietly misinforms, as described in the local search piece.

Common mistakes

Publishing volume. Twenty generated pages about waters you do not fish is twenty pages nobody owns, several of which contain something wrong.

Using it for the personal parts, which are the ones doing the commercial work.

Letting it state a regulation. That is the single highest-cost error available in this trade and it is entirely avoidable by linking out instead.

And assuming that because a tool wrote it, nobody can tell. Readers who fish notice immediately, because generated text about fishing is confident and unspecific in a way real experience never is.

What surprises people

That the position is settled rather than open. The office has published its practice and refused registrations on this basis since before most of these tools existed.

That the reasoning turns on creative control rather than on effort. Spending an afternoon on prompts does not change who determined the expressive elements.

That there is an affirmative duty to disclose generated content in a registration application, resting on a statutory catch-all written long before any of this.

And that the analysis is deliberately fact specific, so a tool used as an editor sits in a genuinely different position from the same tool used as a writer.

The honest answer for a guiding business

Use it to think, to tidy and to check. Do not use it to write anything a client would quote back at you, and do not use it to produce pages you intend to own.

The reason is commercial before it is legal. A guiding business competes on the specific, and the specific is the one thing generation cannot produce. Every hour spent making a machine sound like somebody who fishes your water is an hour not spent writing four sentences that only you could write, which would have done more.

The narrower question of using these tools for a trip report, where the facts come from a day you actually had, is a different case with a better answer, taken up in the trip reports piece. The reply-drafting version is separate again, in the automated replies piece, and the image side has its own rules entirely, covered in the photo tools piece.