The State of Agency Discoverability 2026
How B2B agencies get found now, as referrals thin out and buyers start their search inside AI.
The state of agency discoverability
Momentic Research · May 2026
How B2B agencies get found now, as referrals thin out and buyers start their search inside AI, and what to do about it.
The way agencies win work is quietly changing
What you'll know by the end
Referrals still bring in the most business, and they are slipping. Buyers are starting in AI. Here is what the data shows and where the advantage is for firms that move first.
- Why referrals are cooling. A real decline, with a cause most firms miss: agencies themselves are referring out less.
- Where buyers start now. A growing share open vendor research inside an AI assistant, not a search bar.
- What AI uses to pick agencies. A clear set of signals, most of which live on sites you do not own.
- The cobbler problem. Most agencies do not do for themselves what they sell to clients. That gap is now a cost.
Referrals still win the most work, and they are slipping
The referral cliff
In the RSW/US New Year Outlook surveys, the share of agencies naming referrals their top source of new business fell from 73 percent in 2024 to 58 percent in 2025. That is the steepest drop the study has tracked.
The cause most firms miss is on the supply side. Agencies are using AI to deliver more in-house and protect margin against revenue targets, so they refer out less of the work they used to pass along. Fewer referrals are arriving, and fewer are going out.
73% → 58% — agencies naming referrals their top source of new business, 2024 vs 2025
Source: RSW/US New Year Outlook, 2024 and 2025 surveys.
B2B buyers are opening research inside AI
Where buyers start
Two 2025 datasets point the same way. Close to a third of B2B buyers now begin vendor research in an AI assistant, and most run into an AI Overview before they click anything. The first impression of your firm is increasingly something a model writes.
29% — of B2B buyers start vendor research in an LLM rather than search
- encounter an AI Overview during their research: 72%
Source: G2 2025 Buyer Behavior Report, TrustRadius 2025 (buyer-research vendors).
The same agency search returns four different answers
Four places you get found
Run "best SEO agency for healthcare" through Google, Google's AI Mode, ChatGPT, and Perplexity and you get four results, ranked four ways. A first-place Google ranking now covers one surface out of four.
The signals differ by surface, but the underlying work is shared: real depth, earned authority, and clean structure feed all four. You build it once and show up in more than one place.
- Google organic — the traditional ten links. Relevance, authority, and backlinks.
- Google AI Mode — an AI summary with cited sources. Content depth, structure, and citations.
- ChatGPT — a written recommendation. Knowledge-base presence and recency.
- Perplexity — a conversational answer with sources. Citation networks and authority.
One directory does most of the work
Where AI sources agencies
Across AI-generated agency recommendations, Clutch shows up in about 84.5 percent of citations, far ahead of every other directory. Read that next to who owns it: you do not. Clutch is a third-party surface, and it is where the models go to source agency picks.
This is the clearest example of a broader truth: most of your AI visibility lives on properties you influence but do not control. Profile completeness, review recency, and portfolio depth are the levers you actually hold.
84.5% — of AI-generated agency citations reference Clutch profiles
Source: Hall AI citation analysis (AI-visibility tool).
AEO is mostly influence, not markup
The influenceable web
Schema and clean pages are the entry fee. They are not the engine. Around 90 percent of what AI cites about a firm comes from sources it does not own, and by one count agencies control only two of the seven functions that drive answer-engine outcomes. The rest live in PR, community, and earned media.
The shift is from owning what you publish to influencing what you do not. Treat your own site as the foundation and spend your real effort on the web around it.
- The entry fee: clean Organization, Person, and Service markup; service pages with real methodology; named experts. Roughly a tenth of citations. Necessary, not sufficient.
- The engine: third-party editorial, community presence, reviews, and earned research. Where the other ninety percent of citations come from.
Distribution is the engine, not the publish button
Distribution
If the influenceable web is where citations come from, distribution is how you get there. The web rewards whoever shows up across the most surfaces a model reads, not whoever publishes most. It is easy to copy an idea, and hard to copy a distribution strategy.
- Create once, place everywhere. One real piece of research becomes posts, threads, and pitches across the channels that get cited.
- Spend most of the effort distributing. The work is not the draft. It is getting the draft in front of the surfaces AI pulls from.
- Community is not optional. A large share of "best agency" and "best software" answers trace back to a Reddit thread.
Most agency sites fail the test agencies sell
The cobbler's children
The typical agency site has thin service pages, no real case studies, no named experts, and no structured data — the exact things it bills clients to fix. With a third of buyers starting in AI, that gap is now a measurable disadvantage in a growing channel.
The upside: the firm that fixes its own discoverability first has a story it can actually show, and an advantage most competitors are ignoring.
- What they ship clients: deep service pages, case studies with outcomes, named expert bylines, structured data.
- What their own site has: generic service pages, no published case studies, no named experts, stale and thin content.
AI is reshaping how agencies make money
The business underneath
The public holding companies show where this lands. Margins are up on AI-driven efficiency while revenue wobbles. Production is largely solved, so the value moves to strategy and distribution. The firms caught selling hours in the middle feel the squeeze from both sides.
- Production is commoditized. AI now does the first-draft and audit work. Selling it by the hour is selling the part machines do best.
- Value moves up. Strategy, distribution, and judgment are what clients still pay a premium for.
- Price the outcome. The durable model charges for results, not deliverables. The entry-level tasks are the ones AI absorbs first.
Source: Publicis and IPG investor filings; named SEO Week 2026 stage sessions.
Five moves for agencies
What to do Monday
Ordered by leverage. Each one earns visibility across more than one surface, and each maps to how buyers and AI now actually find firms.
- Fix your own site first. Real service pages, named experts, case studies, clean markup. Be the proof you sell.
- Own your Clutch presence. Complete the profile, keep reviews recent, deepen the portfolio. It is where AI sources you.
- Publish things only you can. Original research and first-party data a competitor cannot copy from a template.
- Distribute, do not just publish. Get that work cited across communities, editorial, and reviews. Create once, place everywhere.
- Sell outcomes, not hours. Move pricing toward results as production keeps getting cheaper.
Be the agency that took its own advice
Where this leaves you
Referrals will keep cooling and buyers will keep starting in AI. The firms that fix their own discoverability, earn presence on the surfaces they do not own, and price for outcomes will compound an advantage while everyone else waits.