Brief
GEO: from dashboards to something an agent can execute
Generative engine optimization is real demand sitting on a noisy vendor market. The moat is moving off the mention tracker and onto content, feeds, and write-back — including shopping.
GEO (generative engine optimization) is the work of shaping brand content and information architecture so ChatGPT, Perplexity, Google AI Overviews and AI Mode, Gemini, Claude, DeepSeek, Doubao, and the rest mention, cite, or recommend you. SEO fought keywords. GEO fights the retrieval and citation weights behind an answer (Yicai).
The demand story is not subtle. People take the synthesized answer and stop. Reported figures disagree on magnitude — one Chinese account said Baidu search volume halved in a half-year; Gartner-style forecasts talk about a 25% cut in traditional search by 2026 — but they agree on direction. The product question is what you sell into that shift.
The living catalog is the GEO map. Shopping-specific GEO sits on the Agentic Commerce map.
Four layers, already blurring
| Layer | Names | Pitch |
|---|---|---|
| Pure monitoring | Peec, Otterly, Rankscale, SE Visible | Mentions, citations, sentiment, share of voice. Often will not write content. |
| Full-stack GEO | Profound, AthenaHQ, Scrunch, GetMint, Yolando | Dashboard plus generation, agents, or a machine-readable site copy |
| SEO-suite modules | Ahrefs Brand Radar, Semrush AI Visibility, BrightEdge, Conductor, Surfer | Bolt AI visibility onto an existing keyword graph and CS book |
| Vertical / enterprise | Birdeye (multi-location), Bluefish (Fortune 500 agentic marketing), The Prompting Company (agent-facing product pages) | A scene, not a generic tracker |
Profound is the only unicorn so far: New York, August 2024, about $155M raised, $1B valuation after a February 2026 C round led by Lightspeed (Khosla, Kleiner, NVIDIA NVentures, Sequoia also in). 700+ customers, company-reported ~10% of the Fortune 500. Agents that draft comparison and social content; claimed 700% lifts in AI-referred traffic. Treat vendor-claimed lifts as positioning.
Peec AI is the large European pure-play: Berlin, 2025, $21M Series A in November 2025 (Singular), ~$29M cumulative, 3,000+ customers (Zalando, Hugo Boss, Chanel). Explicitly: measure and advise, do not write or publish. From $95/month. That anti-position is the product.
Scrunch sells an “Agent Experience Platform”: a structured twin of the site for AI crawlers, human site untouched. $15M Series A (Decibel, July 2025), $26M cumulative. Bluefish is the enterprise OS version of the same job — prompt corpus, multi-model runs, source diagnosis, activation, measurement — $43M Series B in April 2026 (Threshold + NEA), $68M claimed total, customers they name include Adidas, Amex, Hearst, LVMH, Ulta. Company-reported ~10% of the Fortune 500; same caveat as Profound.
Incumbents are not asleep. Ahrefs Brand Radar sits on 468 million monthly prompts and lets you query any brand with no setup. Semrush sells prompt research as keyword research at $99/domain/month, with thinner engine coverage (no Claude/Copilot in the current toolkit). Surfer feeds visibility back into the editor. Their advantage is the installed base, not a prettier share-of-voice chart.
Funding went exponential: a mid-2025 roundup had five names totaling ~$64M, Profound already $23.5M of that; by February 2026 Profound alone was a billion-dollar company. Tallies of “$200M+ into GEO” versus “$31M” disagree mainly on whether Profound, Bluefish, and Peec count. North America ~60% of disclosed rounds, Europe ~40%, with Europeans talking compliance and localization. A vendor blog citing Intel Market Research ($1.01B GEO services in 2025, $1.48B in 2026) is unverified — leave it off the plan.
Pricing is already stratified: $20–$29 monitoring (Rankscale, Otterly), $99–$500 war zone (Peec, AthenaHQ, Semrush, SE Visible), enterprise custom (Profound, Scrunch, Bluefish).
China: real demand, a polluted information market
By June 2025 China had 515 million generative-AI users; trade press says most marketing leads already put “AI visibility” on the annual scorecard. A 940-billion-yuan 2026 GEO market at ~170% growth is a single industry article — do not treat it as a TAM. For scale: China SEO was ~¥100B and SEM ¥300–400B in 2022. There is now a trade-association standard for GEO services in circulation (T/CGCC 119-2026) and talk of a self-discipline pact against “AI poisoning.”
The content about Chinese GEO vendors is itself a GEO case. “Top 10 GEO agencies” lists recycle across Sina, IThome, Tencent Cloud, CSDN, Sohu, Ifeng, almost always ranking the publisher first, with unaudited stats (“86.4% renewal,” “99.9% accuracy”). Relatively checkable names: Zhitui Era (Shanghai, May 2025, 37 Interactive Entertainment in the seed, four-module SaaS, offices in Warsaw, Gothenburg, Singapore); PureblueAI (ex-ByteDance Doubao marketing lead Lu Yang, tens of millions RMB in September 2025, cross-border GEO plus AI-content risk); Zhaixing AI (iFlytek subsidiary money, Spark-based vertical model); Soujixing / InsGEO (Zhuhai Biyou, 12+ Chinese and Western models, “Star Shield” hallucination/poison detection — described mostly in self-promotional posts). Diligence belongs on filings and third-party press, not on the leaderboards.
Five product notes
- Dashboards are the first wave. Execution is the second. Profound Agents, Scrunch AXP, Yolando, The Prompting Company are already generating pages an engine can cite, or surfaces an agent can call. Pure monitoring’s moat is thinning.
- Incumbent SEO suites close the “we built monitoring first” gap. Pure-plays have to win on engine coverage and on an execution layer, not on being early to the mention tracker.
- The $99–$500 band is a bloodbath. Differentiate on a scene (shopping, local, enterprise workflow) or on write-back.
- China is a demand signal and a trust problem. Asymmetry is an opening if the product and the content are more honest than the farms.
- Shopping is the next required module. Peec, GetMint, Profound all moved into “was this SKU recommended.” That is the bridge into Agentic Commerce — and Amazon’s seller-side vacuum is more open than ChatGPT Shopping, where Profound and Peec already sat down.
Bluefish, as a benchmark
Bluefish is not “track ChatGPT mentions.” It is an enterprise loop: large prompt sets, many models, parse answers and citations, prioritize issues, act on content / PR / PDPs / third-party sources, measure again. The useful idea is that a brand’s AI representation is mostly not its own homepage — it is reviews, Reddit, YouTube, publishers, retailer pages, entity data. Strengths: prompt-level evidence, source-influence diagnosis, multi-engine, Fortune 500 governance. Caveats: outputs are probabilistic (methodology and sampling matter); “influence” is not control; revenue attribution is immature; the motion is quote-based, not $29/month.
If you are building in this stack, copy that loop, not the press-release customer list. Cards for all eighteen names: GEO map.