Ask creators to name their biggest problem and the answer is not money. Between 54% and 60% say it is getting their work seen. That was true before this year and it became considerably more true during it.
Something structural happened to discovery in 2026, which means a large amount of the advice available on this subject now describes a system that no longer operates. This page covers what changed, what replaced it, and what to do about it.
Search stopped sending people#
The numbers are not subtle, and they come from four separate measurements that all point the same way.
Sources: SparkToro, Search Engine Land, Search Engine Journal.
There is one exception, and it is the most useful sentence on this page. When someone searches for a specific site or brand by name rather than for a topic, AI Overviews are associated with an 18% increase in click-through.
Topic searches get answered in place. Name searches get sent through. Everything below follows from that.
Where people look now#
About 41% of people rely on AI summaries instead of clicking through, and 13% skip search engines entirely in favour of assistants. Among 18 to 24 year olds, 66% use ChatGPT to find information against 69% still using Google — effectively level for that age group.
Short video continues to carry an enormous share of attention, and social platforms are the main way younger audiences find anything new. Recommendation feeds have not weakened. It is search specifically that stopped working as a route to your work.
The four things that still work#
Being the thing people type in#
Since the only search behaviour that reliably produces a click is someone looking for you specifically, the goal changes from ranking for a topic to being nameable.
That is a content problem rather than a technical one, and it comes down to having a position rather than coverage. “Reviews of kitchen equipment” is a topic anyone can answer. “The person who tests kitchen equipment with arthritic hands” is a name someone remembers and types. The narrower and more specific the identity, the more likely it is to survive in someone’s memory for the week between hearing about you and looking you up.
Practical test: could someone describe you to a friend in one sentence that would let the friend find you? If not, you do not yet have the thing that discovery now rewards.
Appearing on other people’s things#
The most underrated discovery channel available, and the one that scales worst with effort, which is precisely why so few people do it.
Podcast guest appearances, being interviewed, writing a piece for someone else’s newsletter, co-hosting, appearing in someone’s video, answering questions in a community with a name attached. Every one of these borrows an audience that already trusts the person hosting you, which is the transfer that recommendation feeds are worst at.
How to actually get them, since “network more” is not instruction: find shows and newsletters serving the audience you want, at roughly your own scale or slightly above. Not the largest ones. Send a specific proposal rather than an availability notice:
Subject: episode idea, why grip strength ruins most kitchen gadget reviews
Hi. I listen to the show and the episode on adaptive tools in June was the reason I started testing this stuff properly.
I’m a physiotherapy assistant who reviews kitchen equipment for people with limited hand strength. The angle I could bring that you haven’t covered: almost every mainstream review tests with full grip, which is why the recommendations fail for a third of the people reading them. I’ve got specific examples and some numbers from testing.
Happy to send three or four talking points if it sounds useful.
The blank version:
Subject: episode idea, [the angle in eight words] Hi. I listen to the show and [the specific episode] was [the specific effect it had on you]. I'm [what you actually do]. The angle I could bring that you haven't covered: [the claim, stated as a claim]. I've got specific examples and some numbers from [where your numbers come from]. Happy to send three or four talking points if it's of interest.
The second paragraph is the whole pitch. A host is not deciding whether you are impressive, they are deciding whether you will fill forty minutes with something their audience has not heard. Give them the thing their audience has not heard, in one sentence, and let them picture the episode.
A proposal with a specific angle and evidence of having consumed their work is a different object from a pitch, and it gets a different reply rate. Ten of these is more valuable than a month of posting.
Channels you own#
An email list, a community, a podcast feed. The only routes to your audience no algorithm change can alter, and their relative value rose sharply this year simply because everything else became less reliable.
They do not find you new people, that is the trade. What they do is stop you losing the people you already found, which turns out to be where most audiences leak. Owning your audience covers the mechanics, including the technical setup that determines whether your email arrives at all.
Being cited inside AI answers#
If a large share of questions now get answered without a click, the remaining question is whether you are the source being summarised and named. This is real, it is new, and considerably more is being sold about it than is actually known. Being found by AI assistants separates the parts with evidence from the parts that are inference.
What stopped working#
Articles answering common informational questions. This is the format AI answers replace most completely, and it was the backbone of a lot of creator traffic. “How to do X” content now competes with an answer that appears above it and requires no click.
Product comparison content built for search. Same reason, plus it is the affiliate format, which means the traffic and the revenue disappeared together. See affiliate income.
Publishing more of anything. Volume is precisely what platforms began refusing to pay for this year. YouTube’s Inauthentic Content policy names templated mass production, Meta’s originality rules target reposting. Output stopped being a strategy and became a risk.
Keyword optimisation without asking whether a ranking still produces a visit. For topic searches, increasingly it does not. Ranking first on a query that is answered in the overview is a vanity position.
Any guide telling you to publish twenty articles a month targeting long-tail keywords is describing a business that worked until about 2024. It is not that the advice was wrong; the thing it described has gone.
The platforms themselves#
Recommendation feeds still distribute and in some ways matter more now that search does not. But every large platform spent 2026 changing what it will amplify, mostly against anything that looks mass-produced.
There is a discovery consequence people miss here: the formats being demonetised are also being shown to fewer people. A templated channel is not merely earning less, it is reaching less. What gets you demonetised is as much a distribution document as a payment one, and it is worth reading before you design a format rather than after.
A realistic ninety days#
If nobody is finding you, the instinct is to make more. Usually the better move is to make the same amount, differently.
- Weeks one and two. Write the one sentence someone would use to describe you to a friend. If you cannot, narrow until you can. This is the whole foundation and most people skip it.
- Set up an email list and something specific worth joining it for. Twenty minutes for the authentication records, an afternoon for the thing itself.
- Weeks three to eight. Pick one platform that still distributes your kind of work and publish consistently on a cadence you can hold. Not more than you were doing. Better than you were doing.
- Put the list invitation everywhere, and say it out loud rather than only writing it.
- Weeks four to twelve, in parallel. Pitch ten guest appearances with a specific angle each. This is the highest-yield activity in the list and the one you will avoid.
- Publish one thing only you could publish, your own data, your own testing, your own numbers. It is what makes you citable and it is what people remember your name for.
- Throughout. Check metrics weekly, not daily.
The uncomfortable truth underneath all of it: discovery in 2026 rewards being specific and memorable far more than being prolific, and that is a harder thing to do than posting. It is also the only version that compounds.