THE MARKET IS CRACKING. SEO IS DYING. DISCOVERY IS RISING.
In just three years, the channel that underpinned the entire SEO category has ceased to be the same. Why classic SEO stopped being enough, how AI builds an answer today, and how to become the company models cite when buyers ask about your category.
The market is cracking. SEO is dying. Discovery is rising.
Within three years, the ultimate certainty—that Google is the only place where people search—has shattered. Part of the traffic still goes where it always did. The rest is moving to AI: ChatGPT, Perplexity, Claude, Gemini, and AI Overviews. In 2026, AI is so hungry for high-quality sources that entering this space requires a fraction of the budget that will be needed in just two years. This is a window of opportunity. A brief one.
You can have flawless SEO and still remain completely invisible to AI.
What changed
In 2022, ChatGPT proved that searching could be done differently. You do not type keywords; you ask a question. You get an answer—without 10 links, without scrolling. Three years later, hundreds of millions of people use it weekly, including most B2B decision-makers. Perplexity has become the default engine for a growing group of professionals. Claude is embedded in daily business operations. AI Overviews sit above organic results for most categorical queries.
What is missing from this list? Traditional SEO. It has not disappeared—but what used to be the entire playbook is now just the baseline entry requirement, not a winning strategy.
Companies spending $50k+ PLN per month on Performance Max have been reporting the same thing for months: it does not deliver like it did a year ago. AI answers display ahead of ads; Performance Max cannibalizes brand search and remarketing; the algorithm optimizes for conversions that are becoming increasingly scarce.
Paid ads are not an alternative to the problem described in this manifesto. They are the exact same problem—just paid for.
The Polish enterprise B2B market is 2–3 years behind the US. That is not a weak position—it grants lower competition within Polish AI responses, and time to establish category dominance. The first one or two companies to enter in 2028 will fight for crumbs. The first two will divide the cake.
From index to consensus
SEO was built around a single assumption: the search engine is an index, and your job is to rank high within it. Discovery Stack begins with a different premise: AI is not an index—it is a consensus. You do not win by ranking highest; you win by being the answer, because enough diverse sources confirm that you are the correct answer.
Twenty years ago: index · ranking · positions 1–10
The search engine is a library. The goal was to get on page one. Being in the top ten was enough to generate revenue.
Now: consensus · the answer · position 1
LLMs answer in one sentence. Position 10 does not exist. Competition has narrowed tenfold.
A model deems you the correct answer when multiple independent sources say the exact same thing.
SEO ≠ GEO ≠ Discovery Stack
| Method | Scope | Goal | On-site share |
|---|---|---|---|
| SEO | Technical + content + links on your domain | Organic ranking in SERPs | ~100% |
| GEO | Content structure for AI Overviews | Citations in generative answers | ~80% |
| Discovery Stack | Full citation ecosystem | Consensus across AI responses | ~15% |
If someone is selling you GEO or LLMO today purely as on-site optimization for AI Overviews, they are selling you less than one-fifth of the solution. The rest occurs in external mentions, third-party platforms, authority graphs, and brand consensus.
The four pillars
Mentions—uncontrolled references across the web. Earned media—acquired visibility on authoritative third-party sites. Platform-specific—presence in repositories AI accesses directly. Consensus—the absolute alignment of all the above. All four are mandatory. Strong in only two out of four means you are only half effective.
Uncontrolled references
Reviews, posts, comments, lists, audio with transcripts. Models weight statistical associations—200 mentions from 50 locations beat 1,000 from one blog.
Acquired placements
Industry reviews, category lists, podcasts, expert analyses. The authority of the context dictates how heavily a model cites you.
Where AI looks
Encyclopedic bases, thematic communities, B2B expert networks, review platforms, audio with accurate transcripts.
The anchor
Website, founder bio, third-party profiles, press, and audio must say the exact same thing. Models avoid repeating unvalidated claims.
Every company starts from a different baseline. The work lies in auditing where you stand on each pillar and intentionally constructing the missing pathways. That is the core of the Diagnosis phase.
The four workstreams
Discovery Stack is not a single technique. It is a stack of interconnected components organized into four workstreams that must run in parallel. Weak in the fourth, and the other three operate at half their potential.
- Workstream 01 · Measurement. Tracking where AI cites you, where it does not, and where it cites your competition—without a baseline, every move is a shot in the dark.
- Workstream 02 · Production. Website, essays, case files, expert insights—content engineered to be extracted by models as a definitive primary source, not traffic bait.
- Workstream 03 · Distribution. Deliberate presence in environments AI treats as authoritative—encyclopedias, audio, lists, trade press—each with its own logic.
- Workstream 04 · Coordination. Absolute consistency of signals across every surface. Without it, a model perceives four different companies in a single query.
The four workstreams are not sequential stages. They operate simultaneously.
Discovery Stack is a fully managed transformation inside your business—not consulting that ends with a slide deck. It requires a dedicated internal coordinator, typically the CMO, Head of Marketing, or Founder. Collaborative execution, not passive outsourcing.
Empirical proof: Case DSAI-001
The first client of Discovery Stack was its creator—Karol Tabiś at karoltabis.com. Traditional SEO, steady publishing, three years of work—yet virtually invisible to AI while a competitor held a 9× share of voice. Post-implementation: 87 citations in ChatGPT, 90 in Claude, 9 in Google AI Overviews. Q1 2026. Independently measured.
Citation audit (January 2026): zero across ChatGPT, Claude, and AI Overviews. Dominant competitor: 64.3% SOV. Karol: 7.1%—a ninefold advantage bulletproof against traditional SEO alone.
Verification:Open ChatGPT or Claude and ask for experts in Polish B2B sales consulting—see if Karol Tabiś is returned. This strategy does not require you to take the author's word on faith. Read the full case file →
1. A competitor with 9× SOV is not a reason to abandon a category—it is an argument for choosing a different discovery channel.
2. Traditional SEO + standard content marketing does not purchase citation share in AI.
3. Early implementation in 2026 yields compounding returns; waiting until 2028 yields significantly less.
4. What worked for me will not apply 1:1 to your business—the Diagnosis determines where your signals need to be anchored.
A ten-year window
The AI era of discovery unfolds in three phases: The Wild West (2026–2028), Maturation (2028–2030), Monetization (2030+). Those who build during Phase 1 secure digital rent throughout Phases 2 and 3. Those who delay until Phase 2 pay multiples more for the same footprint.
The wild west
Models are still learning which sources to trust. No fixed canons in most Polish B2B niches. Cost of entry: 1×.
Maturation
Category canons solidify. You are buying against lost time. Cost of entry: 3–5×.
Monetization
Paid layers inside AI responses. Companies without an organic stack pay for every appearance. Model: paid only.
Year of start
2026: high chance of default dominance. 2028: work shifts to being included, rarely the top choice. 2030+: paid layer.
| Year of start | Profile | Chance of default dominance |
|---|---|---|
| 2026 | Low costs. Minimal competition. | High |
| 2027 | Costs 2–3× higher. Competitors waking up. | Medium |
| 2028 | Costs 5–10×. Fight to be included. | Low |
| 2030+ | Model mirrors today's Google Ads. | Paid layer |
Convert time while conversion is cheap.
What to do now
Three strategic directions: the next 30 days capture the baseline; 90 days anchor the core; 365 days run the full stack. A concrete roadmap—in full resolution—is the body of work inside Diagnosis and Execution.
30 days · Diagnosis & measurement
- Is your brand turning up in AI responses for your category? Prompt multiple models for top experts and providers in your space.
- Do you exist structurally outside your website? Cross-reference search engines, LLMs, and citation analytics.
- What does your competition have that you lack? Catalog whose names keep appearing when AI answers category questions.
90 days · Core foundations
- Format your website and expert assets so models can parse and extract them as authoritative sources.
- Align messaging across homepage, executive bios, and major third-party profiles into a single narrative.
365 days · Full-stack scaling
- Quarterly citation velocity reviews across major LLMs.
- 2–4 deep authority assets published per month.
- Programmatic placements on category source nodes.
- 6–12 week synchronization cycles across all external signals.
In business, the winner is not the one who relies on hope. The winner is the one who relies on data.
Who this is for
Discovery Stack is engineered for B2B enterprises with long sales cycles, organizations of 20–200 people, and teams ready to act as strategic partners—not passive clients.
This is for you if
Buyers research 1–12 months before buying. Premium offering (LTV >50k PLN). Validated product. Performance Max no longer performs like it used to. Ready to partner in the process.
This is not for you if
B2C high-volume low-value. Local consumer services. No budget for a year-long track. Looking for a quick plugin fix.
The three control filters
- Lift in Direct traffic in GA over 6 months without an obvious source? (LLM traffic often reports as direct.)
- AEs hearing less "I'll Google you" and more "I'll look into you"?
- Willing to allocate budget to an emerging discovery channel for default AI citation status in 12–18 months?
3× yes → let's talk. 2× yes → strategic fit. 1× yes → timing not right yet. 0× yes → reconnect in a year.
Frequently asked questions
How is Discovery Stack different from SEO?
SEO optimizes your site for search engine rankings. Discovery Stack builds your company's citability in AI answers—where about 85% of cited sources come from outside your site. SEO is today's entry ticket, not the winning strategy.
How is Discovery Stack different from GEO and AEO?
GEO and AEO focus on on-site structure for AI answers. Discovery Stack covers every source AI uses to build consensus—mentions, earned media, platform presence, and signal alignment. GEO is a fragment; Discovery Stack is the whole.
Who is Discovery Stack for?
B2B companies (20–200 people) with long sales cycles where buyers research before deciding—and increasingly ask AI. Minimum LTV around 50k PLN per client and readiness to partner in a guided change process.
When do you see first results?
Discovery Stack is work measured in years. Foundations in the first 90 days; full stack within a year. In 2026 models are hungry for good sources—early implementation compounds faster than waiting until 2028.
Is the manifest paid?
No. The full manifest is public and free—including the PDF. Paid engagements begin with Diagnosis and the Execution program, where we deliver a company-specific implementation map.
Sources & empirical base
- infoShare 2026 (Gdańsk)—+600% Google CPC in the US; Performance Max decay among Polish advertisers at 50k+ PLN monthly budgets.
- AI Citation Audit: Case DSAI-001—measurements 2026-05-25; publicly verifiable by querying models on Polish B2B sales consulting.
- Share of Voice analytics—Polish B2B sales consulting, Q2 2026: 64.3% leader vs 7.1% pre-stack.
- McKinsey, Gartner, Forrester (2025–2026)—AI footprint across enterprise buying paths.
- Brand mentions vs backlinks in AI citation share (2025–2026)—unlinked context-rich mentions as stronger predictors than classic links.
- RAG architecture and training source weightings—public technical documentation.
Specific links and full report titles are shared during individual consultations.
Next step · qualification consultation
30 minutes. Complementary. With me directly.
We evaluate whether Discovery Stack fits your business model, which pillars hold your critical gaps, and whether a full Diagnosis is justified. We map authoritative platforms for your category—mechanics this manifesto keeps confidential.
Schedule your consultationDS-MANIFESTO-001 · MAY 2026 · DOCUMENT ENDS
Signed · Karol Tabiś · Founder, Discovery Stack