September 23, 2026
September 24, 2026
Editorial Disclaimer
This content is published for general information and editorial purposes only. It does not constitute financial, investment, or legal advice, nor should it be relied upon as such. Any mention of companies, platforms, or services does not imply endorsement or recommendation. We are not affiliated with, nor do we accept responsibility for, any third-party entities referenced. Financial markets and company circumstances can change rapidly. Readers should perform their own independent research and seek professional advice before making any financial or investment decisions.
Most advice on winning agency clients is a list of things to add: another channel, another ad retainer, another outreach sequence. It feels productive, and for a while it probably is.
The trouble is that nothing ever comes off the list. Channels that stopped paying two years ago still get their monthly budget and their Friday-afternoon hours, because nobody made a decision to end them. If you've ever looked at your marketing spend and struggled to say which line brought in the last three clients, you already know the feeling.
So we asked nine people who run marketing companies a different question: what's one way of getting customers you stopped this year, and what replaced it? Their answers are below, in their own words. If you want the fundamentals first, choosing a niche, capturing leads, responding quickly, our guide on how to grow a digital marketing agency covers them. This piece is about what to stop.
For years the chain was simple: rank for the question, get the traffic, get the enquiry. This year the first link held and the last one didn't.
There's independent data behind that. When the Pew Research Center tracked the browsing of 900 US adults, it found that people clicked a traditional search result on 8% of visits when an AI summary appeared, against 15% when it didn't. That is awkward to spot from inside a business, because the dashboards most of us check first still look healthy. Joe Davies at fatjoe saw it in the leads, not the rankings.
This year we stopped treating top-of-funnel SEO content as our main acquisition channel. Informational traffic was still coming in, but AI Overviews and ChatGPT were answering the question before anyone reached us, so the leads dried up even as rankings held. We replaced it with founder-led content and paid creative, mostly me on LinkedIn talking directly about what's changing in search, backed by Meta ads built around the same points of view. The result was fewer visitors but far more qualified conversations, and a pipeline that no longer depends on Google deciding to send us a click.
Joe Davies, Founder and CEO of fatjoe
Check this in your own numbers before you change anything. Take your ten best-ranking informational pages and count the enquiries each one produced last quarter. Count enquiries, not impressions. If the answer is close to zero, those pages are doing a brand job now. They may be worth keeping, but your pipeline shouldn't rest on them.
Three of the nine answers end up in the same place. The channel they dropped was different each time (newsletter ads, Facebook, paid social), but the budget and the effort moved towards the same goal: being one of the sources an AI tool reaches for when a buyer asks who to hire.
Some budget lines never fail outright. They return a little less each year, and because the cost is small nobody gets round to questioning it. Chilli Fruit's newsletter advertising was one of those.
For five years we put around $8,000 a year into newsletter ads. It never dented the budget, but the results got weaker every year as the newsletter space filled up with dozens of titles that nobody reads. For 2026 we dropped it completely and moved the budget into AI visibility: publishing content, backed by real data, testimonials and awards, that ChatGPT, Gemini and Claude pull from when someone asks who the best SEO consultants are. It didn't exist as a channel two or three years ago and it's now roughly half of all our leads. And unlike a newsletter ad, it doesn't disappear after a month.
Milosz Krasinski, Managing Director, Chilli Fruit
A useful test for any recurring spend is to put this year's result next to what the same money bought three years ago. A small line that has slid for years is easy to defend, because it's cheap, and rarely worth defending. Look at what Milosz says the replacement runs on, too: real data, testimonials and awards. Gather that proof before you move the budget.
Milosz's version of AI visibility is publishing. Mike Khorev at Nine Peaks Media went at it from the other side, by getting the agency mentioned in the places AI tools already treat as sources. His list isn't guesswork. Ahrefs tracks the domains ChatGPT cites most, and Reddit sits in first place with 16.8% of citations at the September 2026 update.
This year we stopped Facebook marketing and moved that budget into getting Nine Peaks Media mentioned in the listicles, Reddit threads, and directories that ChatGPT, Perplexity, and Google AI Overviews use as sources for prompts like "recommend the best AI SEO agency". As a result, more of our new inbound leads now tell us they found us through an AI search, and those leads usually come in better informed and closer to a buying decision.
Mike Khorev, Managing Director at Nine Peaks Media
You can find your own worklist in ten minutes. Ask ChatGPT, Perplexity and Google the questions a prospect would ask, the best agency for your service, in your sector or your area, and note which pages get cited. If you're absent from all of them, those listicles, threads and directories are where to start.
Paid social is easy to switch on and hard to audit, which is how it becomes a default. Kyttn audits ad accounts, and Jake Siddall says almost every one they looked at this year showed the budget bleeding in the same place. Jake calls the replacement deliberately boring.
We stopped treating paid social as a default acquisition channel this year and wound that spend right back, because almost every ad account we audited showed that's where the budget was bleeding. What replaced it was getting deliberately boring: building the SEO layer properly, which now doubles as AI visibility, so the same work gets clients found in Google and cited in AI answers, and the customers arriving through AI recommendations convert at a higher rate than social ever managed. For one salon client, that shift helped cut cost per acquisition from NZ$70.70 to NZ$47.52 and made them first in AI visibility across their local competitors.
Jake Siddall, Founder of Kyttn
Before you touch your own ad spend:
The salon figures in Jake's answer are for one client, so treat them as an example of what's possible and not as a benchmark.
Two answers are about the press, and both swap asking for a favour with being useful to someone who is already looking.
Cold guest-post pitching asks a busy editor for something. Sheraz Ali watched his response rates drop and decided this year that the hours no longer justified the reward. His replacement turns the request around: on HARO and Qwoted the journalist posts what they need, and sources reply.
This year, I stopped sending cold guest posts to editorial email lists. I gave up on that. The response rates dropped and the time required to achieve it was not worth the reward. Now I answer journalist queries through HARO and Qwoted. This has put our clients in the path of writers looking for sources and the published articles land in strong publications all with low volume outreach.
Sheraz Ali, Founder, HARO Links Builder
If HARO sounds like old news, it nearly was. Featured.com bought it from Cision in April 2025 and brought it back as a free service delivered by email. Try a month of answering requests alongside your usual outreach and compare placements per hour. It needs someone checking queries every day and replying quickly with something quotable.
Eight of these nine changes were choices. This one wasn't. C.S. Lewis and Co. Publicists relied on publishing houses to refer their authors, and several of those houses have downsized or closed. David Lewis's answer was to use the firm's own service on itself.
We stopped counting on referrals from publishing houses this year, because several of them that used to send us authors have downsized or closed. We replaced that pipeline by doing for ourselves what we do for clients: pitching ourselves as sources on book marketing and publishing to trade and business outlets, so authors find us through the coverage rather than through a shrinking network of publishers.
David Lewis, Vice President, C.S. Lewis and Co. Publicists
A referral source can vanish for reasons that have nothing to do with how good you are, so it's worth knowing how many of your enquiries trace back to the same two or three introducers. The quickest replacement is usually the thing you already sell. If you get clients quoted in the trade press, put your own people forward the same way.
An RFP is a lot of unpaid senior time. Loopio's 2026 benchmark report puts the average response at 33 hours and the average win rate at 39%. Bryan Sekine found RFPs rarely turned into customers for his agency, so he stopped filling them in.
One way that I stopped getting customers this year was filling out RFPs (Request For Proposal). It was extremely time consuming and rarely converted into a new customer for my agency. Instead, we doubled-down on improving our local marketing, our local SEO, and putting more localised content on our website. As a result of doing those things, we have already secured 3 new local customers that we would have otherwise never known about!
Bryan Sekine, Fox and Owl Marketing
Add up the hours that went into your last five proposals and how many you won. If the maths is poor, spend the next proposal's worth of hours on local search: location pages, content written for the places you serve, and the listings that back them up.
Bryan's three new customers were ones the agency would otherwise never have known about. An RFP only reaches you if someone already has you on a list.
The last two answers aren't about channels at all. Both businesses kept publishing. What changed was the kind of thing they put out.
Opening with a statistic feels safe, because nobody can argue with a number. Villain Branding led almost every post that way, and its newsletters were, by Lauryn Warnick's own description, academic and data heavy. Then the team looked at how those posts performed.
We stopped opening our content with data. A stat used to lead just about every post and our newsletters were quite academic and data heavy. When we checked our own numbers, one data-led post pulled about a third of our average engagement; a personal, belief-led post pulled nearly five times average. Now, we focus on the story we're telling and we prove it with that data that shows up once the reader's already engaged.
Lauryn Warnick, CEO and Founder, Villain Branding
This is cheap to test. Take your last ten or twenty posts, mark each one as opening with a number or with an opinion or story, and compare engagement. Lauryn's comparison is between individual posts, so run it across more than two of yours before you change house style. She didn't drop the data, either. It moved later in the piece, where it backs up a point the reader already cares about.
Generic outbound and waiting for referrals have the same weakness. The prospect arrives knowing almost nothing about how you work, so the first call is spent proving you can do the job. Now Media changed two things at once.
This year we moved away from generic outbound prospecting and waiting on referrals, and focused instead on putting our thinking in public: opinions, frameworks and the reasoning behind the work, rather than just a portfolio of finished projects. Prospects now come into the first conversation already understanding how we think, so it's less about convincing them we can do the work and more about whether we're the right fit. We also handed prospect qualification to AI instead of spending hours on calls working out budgets and scopes, which means the team only talks to people who already know our service, our cost and the likely timelines. That's helped us 10x our conversion rate and our revenues.
Nithin Koshy, CEO, Now Media
The first half costs nothing to try. Next time you would post a finished project, post why you made the decisions in it, including the ones you'd make differently now.
The second half doesn't have to start with AI. A short form that asks about budget, scope and timing before anyone books a call does the same filtering. The 10x is Nithin's own figure for his business, and the mechanism is the useful part.
Nine answers don't make a rulebook, but they do show where to look. Go through your own list of channels and look hard at anything that:
Hold on to anything that:
You don't need to replace everything you cut. Robin's piece on creative ways to attract customers makes the same case from the other direction: fewer, sharper tactics tend to beat volume.
None of the nine added a channel to an unchanged workload. Each stopped something first, whether a budget line or a sales process, and spent what that freed up elsewhere.
Joe Davies ends his answer with a pipeline that no longer depends on Google deciding to send a click. Several of the nine are still investing in search, so don't read that as a case against SEO. Read it as a prompt to work out which platform or referrer could cut off your enquiries by changing its mind.
Pick the channel that produced less this year than last, and decide what would replace it. If you'd like to see how other owners have worked through decisions like this, Robin's case studies are a practical place to start, and if you want help deciding what to drop, get in touch.
Because an AI summary can answer the question before the reader reaches your page. Pew Research found people clicked a traditional result on 8% of visits when an AI summary appeared, against 15% when it did not, so traffic dashboards can look healthy while the pipeline dries up.
It means being one of the sources an AI tool reaches for when someone asks who to hire. Start by asking ChatGPT, Perplexity and Google the questions a prospect would ask, then note which pages get cited. Those listicles, threads and directories are your worklist.
It depends on your win rate. Loopio's benchmark puts the average response at 33 hours and the average win rate at 39%. Total the hours behind your last five proposals and compare that against what the same time spent on local search might produce.
Test it on your own posts rather than assuming. One contributor found a data-led post pulled about a third of their average engagement while a belief-led post pulled nearly five times average. The data still mattered, it just moved later in the piece.
Every one of them stopped something before starting something else. None added a channel on top of an unchanged workload, which is why the freed budget and freed hours had somewhere to go.
