Skip to content
Cited logo

Cited

Cited is a Moldova-based GEO agency for B2B software and AI companies with $2 million or more in annual recurring revenue. It measures which products ChatGPT, Gemini, Claude, Perplexity, and Google AI Overviews recommend in a client's category, fixes site and entity data, and earns third-party coverage, with packages from a $2,500 audit to $9,000 a month.

GEO specialist for B2B software
Cited homepage

Facts

Websitethegeoagency.co
HeadquartersMoldova (agency site)
Founded2025
OwnershipFounder-led
StanceGEO-native
Typical buyerB2B software and AI companies with $2 million or more in annual recurring revenue
DeliverySmall remote agency, with one client per software category
VerticalsB2B SaaS
SpecialtiesB2B SaaS, Startups / SMB

Buyers, use cases, price, and similar agencies

Who uses itIts clients are B2B software and AI companies with at least $2 million in annual recurring revenue, selling in the United States and Europe. Its published results come from ClippingStars, a creator-services brand, and it lists OneShot.ai and Kirgo as engagements with results still in progress.
What forTeams hire it for a baseline of what AI engines recommend in their category, technical and entity fixes so AI can read their site, and comparison content, community work, and press that earn third-party mentions, with monthly tracking of 75 or more buyer prompts.
PriceCited is in the $$ band, the second of our four monthly cost bands. 28 of the 79 agencies we review are in that band. It publishes a $2,500 audit, a $5,500 one-time Foundation Sprint, a Growth plan at $5,000 a month, and an Authority plan from $9,000 a month, with a six-month minimum on the monthly plans.
Similar agenciesTripledart, Embarque, and Daydream are also listed under B2B SaaS in our GEO agency reviews. All three are in the $$ monthly cost band.

GEO / AEO philosophy

Cited holds that a site AI can read is the starting point, and that recommendations come from what trusted third parties say about a product. Its method runs in four stages: a baseline of what AI recommends in the client's category, technical and entity fixes, earning sources AI already trusts, and monthly tracking of the same prompts.

The agency's about page says, "Readiness isn't recommendation. Schema and llms.txt make you readable; they don't make you recommended." It runs each buyer prompt three times on each of five engines because AI answers vary. Its one published result is a Readiness Score for ClippingStars that rose from 41 to 88 out of 100 between January and October 2026, and the agency says that score measures how ready a site is to be read and quoted, not how often AI recommends the brand.

Offering

  • Tests 25 buyer prompts across five AI engines, three runs each, and reports who AI recommends instead of the client and which sources it relies on.
  • Fixes crawler access and rendering, adds structured data, sets up entity profiles on Wikidata, G2, Capterra, and Crunchbase, and restructures product, pricing, and comparison pages so AI can quote them.
  • Writes GEO articles and comparison and alternatives pages, builds disclosed Reddit and community presence, and pitches trade press, with monthly tracking of 75 buyer prompts on the Growth plan.

Editorial scores

We scored the agency on eight dimensions to show a buyer where it is strong and where it is thin. Scores run from 1 to 10. The last column shows each score's difference from the median for all reviewed agencies.

DimensionScoreVs median
Technical GEO61
Lead generation42
Thought leadership5=
Entity / knowledge graph62
Local / multi-location12
Enterprise readiness24
Innovation5=
Measurement72

Service mix

We estimated how the agency's work splits across service types, which shows a buyer where its effort goes.

Measurement 10%Digital PR 15%Content 25%GEO / AEO 35%Classic SEO 15%

Best At

  • a dated baseline of what AI engines recommend in a software category
  • technical and entity fixes that make a site readable to AI
  • comparison pages, community work, and press for B2B software companies

Poor Fit When

  • you need a guaranteed placement in AI answers, which the agency says no one can promise
  • the goal is local, consumer, or e-commerce search
  • you need a month-to-month plan, since its monthly packages carry a six-month minimum

Clients

NameStatus
ClippingStarsNamed in case study
OneShot.aiNamed in case study
KirgoNamed in case study

Innovation

Cited reports two measures of its own design. The Readiness Score rates how ready a website is for AI to read, trust, and quote, and a second rate counts how often AI recommends a client across tracked prompts. Every baseline is dated, and the agency publishes an after score only once it has been measured.

Pricing

Cited publishes its prices as of October 2026: an AI Visibility Audit at $2,500 one time, credited to the first month when a package is signed, a Foundation Sprint at $5,500 one time, Growth at $5,000 a month, and Authority from $9,000 a month. Growth and Authority carry a six-month minimum and then run month to month. Tier-1 press is an add-on at $4,000 to $8,000 a placement.

Pricing vs Performance Spectrum

GEO Agency Price Versus Performance, 2026
GEO Agency Price Versus Performance, 2026Each point is one directory agency. Vertical position is the mean of eight editorial scores. Horizontal position is a monthly engagement cost band from dollar to four dollars. Published midpoints appear in tooltips when available. Default labels mark score outliers only; other names appear on hover.45678Average Editorial Score$$$$$$$$$$Typical monthly engagement costCited

Practice timeline

2025

Lorenzo Galateanu founds Cited.

January 2026

Cited audits ClippingStars and records a Readiness Score of 41 out of 100.

February 2026

Cited records baseline scores of 43 for OneShot.ai and 32 for Kirgo.

October 2026

ClippingStars reaches a Readiness Score of 88 out of 100 in Cited's audit.

Swipe left or right to see the full timeline.

Agency Leadership

NameTitleBackground
Lorenzo GalateanuFounderStudied at SSBM Geneva and has built and run content agencies, AI automation systems, and fintech products. He leads the agency's GEO method.

Evidence

TypeDescriptionSource
Figures reported by the agencyClippingStars' Readiness Score rose from 41 to 88 out of 100 between January and October 2026, in the agency's own site audit.Cited ClippingStars case study
Method and productThe Readiness Score, which rates how ready a website is for AI to read and quote, and a free AI Readiness Check on the agency's site.Cited method page

Related pages