When Your Docs Stop Being the Front Door of Your Funnel

This is not a happy story, and it should be read slowly.

Adam Wathan, who created Tailwind CSS, described in January what AI search had done to his business. Traffic to the Tailwind documentation was down about 40% from early 2023, even though the framework was more popular than ever. Revenue was down close to 80%. He had to let go of 75% of his engineering team — four people down to one. As he put it, the docs were the only way people found out about the commercial products, and without customers the project cannot afford to maintain the framework.

He was writing this in a thread about whether to add an llms.txt feature. His answer was that he wants to figure out how to offer LLM-optimized docs without making the situation worse, and that he could not prioritize it.

That is the shape of the problem in one sentence. The thing that used to bring people in is now the thing that gets absorbed, and the team that would have built the adaptation is the team that just got smaller.

The mechanism, separated from the outcome

For most software and SaaS businesses, the docs, help centre or resource section was quietly the top of the funnel. Someone has a problem, they search, they land on your explanatory page, they learn that you understand their problem, and eventually they buy. The page did three jobs at once: it was the acquisition channel, the credibility proof, and the internal reference.

AI answers took the first job. Not all of it, but the most valuable slice — the short, factual, “how do I do X” questions that a person can now have answered without visiting anyone’s site. What remains is the second job, credibility, and the third, reference. Both are real work. Neither is a growth channel.

This is a structural change, and it is not specific to developer tools. Any business whose marketing was “we publish the best free answer to the question our buyers ask” is in the same position. Local services, legal, health, finance, education — the pattern is the same shape, arriving at different speeds.

What people get wrong when they hear this story

The wrong conclusion is “docs are dead, stop writing them.” The docs are how you stop being a black box to both a human and a model. Drop them and you lose the material that anyone — a customer, a journalist, or an assistant — would cite when explaining what you do.

The second wrong conclusion is “just add the new file format.” Tailwind is, in a sense, the counterexample to that: the problem was not that AI couldn’t read the docs. It read them fine. It just didn’t need to send anyone to them.

The third wrong conclusion is the one that is most tempting for our audience: telling yourself this proves backlinks were the real answer all along. It doesn’t. What it proves is that a business needs acquisition channels it does not own. Search was never owned. Neither is AI. Both are borrowed distribution, and borrowing terms change without negotiation.

What the story actually implies about off-site work

Look at where the value moved. Before, three jobs were done by one page you controlled. Now the acquisition job has moved to a surface you do not control, and the credibility job is being spread across places you only partly control: reviews, comparisons, directories, community threads, partner pages.

That has a specific implication for how you spend. On-site pages still establish what is true. Off-site pages are where a claim gets repeated, and repetition is what these systems use when deciding who is a plausible answer. A product described on forty third-party pages in compatible terms is much harder to leave out than a product described only on its own site, however good that site is.

That is not a pitch for volume — the last year of Google announcements made volume the most exposed strategy there is. It is an argument for treating off-site presence as infrastructure rather than as a campaign, because it is the only part of your distribution that does not depend on a platform deciding to keep sending you traffic.

A short list worth acting on this month

The fifth row is the one most teams fail, and it is the one that would have mattered most to a business in this situation. Not because diversification prevents a platform from changing, but because it buys you the quarters you need to respond.

That is the dull version of the answer, and it is the one that survives platform changes: https://backlinksubmit.org takes a product URL, matches it to relevant directories and communities from a handpicked list, submits manually with written copy, and shows in a console which platforms have been submitted and how each is progressing.

FAQ

Does AI search hurt businesses that publish free documentation?

It has, measurably, for at least one widely publicised case: a 40% decline in docs traffic since early 2023 alongside a steep revenue fall. Whether the effect is as severe elsewhere depends on how much of your audience arrives through short factual queries.

Should we stop investing in documentation?

No. Documentation is how you become legible to customers and to language models alike. The point is that you can no longer treat it as an acquisition channel in your plan and expect the old return.

What’s the replacement for the traffic?

Two things: being named in places you don’t own — reviews, comparisons, directories, communities — so the recommendation can be made without a click, and building channels that are not search at all, such as integrations, partnerships and direct relationships.

Is this an argument for buying backlinks?

It is an argument for having distribution that persists when a platform changes its mind. Placements that describe you accurately on pages with their own audience do that. Bulk link volume does not, and this year Google spent its time making that clearer rather than less.