Retrieval Authority Consultancy
Discovery shifted from pages to answers. We engineer the sources those answers cite.
Buyers no longer scan ten blue links. They ask a model and accept a synthesized answer. We build the durable, citable presence — across Reddit, LLMs, and search — that determines whether your category gets framed around you or around someone else.
The Shift
Search is collapsing into retrieval. Your category is being framed without you.
ChatGPT, Perplexity, Gemini, and Google's AI Overviews increasingly answer the questions buyers used to type into a search bar. The answer is composed from a small set of sources the model trusts — and Reddit is now one of the most cited sources on the open web.
If your brand isn't present in those source threads, you aren't losing a ranking. You're being written out of the answer entirely.
We engineer the discourse, citations, and framing that determine which brands models reach for when a buyer asks.
Search era
- Ten blue links
- User chooses the source
- Rank for the query
- Optimize pages
Retrieval era
- One synthesized answer
- Model chooses the source
- Get cited in the answer
- Optimize the corpus models read
+527%
YoY growth in AI-referred sessions
40.1%
LLM citation share held by Reddit
1 of N
Sources a model picks per answer
Failure Modes
Three ways the old playbook fails.
Each one produces activity. None, in isolation, produces compounding authority. The boundary between them is where value now lives.
Impressions that terminate.
The pattern
Top-of-funnel spend. Paid social, display, influencer seeding. Optimized against CPM, reach, brand recall.
Why it breaks
Impressions leave no indexable artifact. The buyer asks ChatGPT later — and the answer draws from someone else's discourse.
Test for it
Ask five assistants: "Who are the leading vendors for [your ICP's job]?" If your brand never appears, your recall isn't showing up in retrieval.
<1 day half-life for paid search.
Rank without survivability.
The pattern
Content libraries optimized to rank on-domain. Programmatic SEO. Keyword-targeted blog posts. Schema, topical authority, internal linking.
Why it breaks
Ranking a page is not the same as being cited. AI systems prefer corroborated claims across sources. A vendor's own page is the weakest evidence.
Test for it
Count your citations in AI answers for ten queries your best customers run. If owned content is the only source, the surface is shrinking under you.
40.1% of LLM citations pull from Reddit. Owned content barely registers.
Trust the index can't see.
The pattern
Active presence on LinkedIn, X, Instagram. Communities on Discord, Slack, gated Substacks. Founder content at scale.
Why it breaks
Most social surfaces are poorly indexed or walled off entirely. LinkedIn posts are rarely cited. Discord and Slack are invisible to retrieval.
Test for it
Map your top ten social posts against AI citations for adjacent queries. If none surface, the trust you earned is invisible to the system.
90% of community audience is lurkers. Visible engagement is a 1% sample.
The Method
How we operate.
Four phases. Every phase produces durable artifacts. Phase 01 happens once. Phases 02 through 04 run continuously — retrieval authority is infrastructure, not a campaign.
We measure where you currently appear in AI answers, map the discourse surfaces that matter for your category, and score every existing asset for decay risk. This is the only phase that ends. The rest run continuously.
What happens:
- •Retrieval share-of-voice baseline across five AI assistants
- •Competitive citation analysis: who out-retrieves you and where
- •Discourse surface mapping (Reddit, forums, review ecosystems, practitioner communities)
- •Decay risk scoring of existing content and citations
- •Prioritized remediation roadmap
You know exactly where you are invisible, where rivals are winning, and what to fix first.
We design the discourse: which threads, by whom, under what identity, in language that will compress cleanly when an AI summarizes it. Category framing decisions get made here. Research pipeline gets scoped.
What happens:
- •Discourse architecture: threads, identities, posture per community
- •Category framing document — phrases and definitions the brand will own
- •Compression-tested messaging (claim fidelity at 200 / 50 / 15 word summaries)
- •Research pipeline scoping — what we will publish under our name
- •Per-community voice and engagement guardrails
Every contribution is designed before it ships to survive both community scrutiny and algorithmic compression.
Continuous discourse work on the surfaces that feed retrieval. Original research published on a quarterly cadence. Every deliverable passes compression stability QA before it ships.
What happens:
- •Thread participation across mapped Core and Adjacent communities
- •Original research publications, signed and dated
- •Supporting asset creation on cadence (newsletters, explainers, comparisons)
- •Pre-publication compression QA: four assistants, three compression ratios per claim
- •Cross-source corroboration to close attribution gaps
Durable retrieval assets that get cited for quarters, not days.
Retrieval relevance decays on a schedule. A thread that was definitive in 2024 loses ground to a mediocre but active one in 2026. Maintenance is the job, not an afterthought.
What happens:
- •Monthly retrieval share-of-voice tracking and reporting
- •Quarterly decay checks on every citation-source asset
- •Thread revival when citation share slips
- •Source refresh — dated updates so newer corroboration exists for the same claim
- •Competitive monitoring and response when rivals move
Authority that compounds across quarters instead of decaying after the campaign ends.
About Reddit Ads
We can run Reddit ads, but they're not the main thing we do.
Ads can help, but only after you've built some credibility. Running ads without community presence is a waste of money on Reddit.
When ads make sense:
What we don't use ads for:
When we use ads, they're for:
What We Won't Promise
Retrieval authority compounds. It doesn't shortcut. We name what this work isn't, up front.
If you need a quarterly pipeline number tied to a single channel, we're the wrong partner. If you want to own the way your category gets answered, keep reading.
Fit
Who this is for.
A specific shape of company. We say no to most inbound.
Strong fit
Poor fit
Who This Works For
Reddit works best for products people talk about and compare.
Good fit
- B2B software companies
- Developer tools
- Healthcare and wellness (education-focused)
- Fintech and financial services
- AI and tech infrastructure
- Products people research before buying
Not a good fit
- Consumer brands chasing viral moments
- Companies that can't handle public criticism
- Anyone wanting results in 30 days
- Products that rely on hype over substance
What We Measure
Retrieval Share of Voice. The metric that replaces impressions in an answer-first internet.
Impressions told you who saw a page. RSoV tells you which brands models reach for when a buyer asks — and how often the answer is framed in your terms instead of a competitor's.
Standard metrics we still track
Retrieval Share of Voice — the dimensions we report
Citation Share
Share of mentions when leading LLMs answer category questions.
Framing Share
Whose category definitions, comparisons, and criteria the model adopts.
Source Composition
Ratio of owned, earned, and community sources behind each answer.
Competitive Delta
Your retrieval position versus the named competitive set, tracked over time.
Compression Survival
Which assets survive when models compress long threads into a single answer.
Decay & Revival
Half-life of cited sources and the cadence required to keep them live.
Durable Retrieval Assets
What you own after 6 to 12 months of disciplined retrieval work:
- →Your brand named in LLM answers when buyers ask category questions
- →Reddit threads you seeded ranked on page 1 of Google for high-intent queries
- →Your framing — your criteria, your comparisons — adopted by models and buyers
- →Community sources that keep getting cited 6 to 18 months after they're published
- →A retrieval position that compounds instead of resetting with every campaign
Cited sources have a half-life of 6 to 18 months. Engineered well, they keep paying out long after the campaign closes.
That's the asset. Not impressions. Not upvotes. Position inside the answer.
Before you book a call
Fit, timelines, and how we actually measure Retrieval Share of Voice.
The questions every serious buyer asks in the first call. Answered here so the call can be about your category instead.
Companies with a defined category, a named competitive set, and a 6–18 month horizon. You should already have product–market signal and care about being the answer when a buyer asks an LLM — not about hitting a pipeline number this quarter. If you need short-cycle leads or vanity reach numbers, we are not the right partner.
Want to see if Reddit makes sense for your brand?
We'll look at your product and target audience and tell you whether Reddit is worth pursuing. No pitch, just an honest assessment.