Buyer tools
The GEO agency RFP: 25 questions, and who answers them in public
Of the 25 questions a buyer should put to a GEO agency, nine draw no Yes from any of the ten agencies we read, and the median agency answers 5 of 25 on its own site. None of them shows the raw answer text behind a result, and none lets a buyer pay without a call.
Broadcastwell publishes this list and appears on it, at number one. It is an editorial list scored on criteria a buyer can verify, not a measured ranking. The scoring, the sources and the dates are below.
What the count shows
Key findings
- 9 of 25questions draw no Yes from any of the ten agencies read, and two of them draw not even a Partial.
- 5 of 25is the median number of questions an agency answers in public. Citant answers the most, 12 of 25.
- 0Yes answers across all ten agencies on the four evidence questions, Q7 to Q10: raw answer text, cited sources per answer, variance and repeatability.
- 0 of 10agencies let a buyer pay without a call.
- 20 of 25is Broadcastwell's own score on the same rule. The five questions it does not fully answer are listed below.
How this page was scored
Ten GEO agencies and Broadcastwell, each read against the same 25 questions on its own public site. A question counts as answered only when a buyer can find the answer without a call, a sign up or an email.
Yes
The company's own site answers the question in full. Every Yes on this page links to the page that answers it.
Partial
Part of the answer is published, for example a cadence where the question asks for a run count. Partials are printed but never added to the count.
No
No answer was found on the pages read. That is a claim about the pages, not about how the company works.
Read date
Every cell was read on the company's own public pages on 17 September 2026. No sign up, no email, no call.
Three limits carried from the read: Minuttia's method answers come from a published data study rather than its service pages, and the same applies to parts of Omniscient Digital's and Siege Media's; iPullRank's AI Search Manual sits behind an email form, so it was judged on its ungated pages only.
One cell was re-read on 17 September 2026: First Page Sage's price question was changed from Yes to No, because the tiers on its cost page are framed as prices seen among other companies rather than its own rates.
The 25 questions
Five groups of five. Each question says why it matters and what a verifiable answer looks like, with how many of the ten agencies answer it in public.
Method, Q1 to Q5
Q1. Is the question set published?
Why it matters. The questions decide the answer. A score built on questions you cannot read cannot be checked or repeated.
A verifiable answer. The actual questions, or the rules they are written to, printed where a buyer can read them before paying.
Q2. Are the classification rules published?
Why it matters. Whether an answer counts as naming you depends on the rule. Loose rules inflate a score.
A verifiable answer. A written rule for what counts as named, cited or absent, for example word boundary matching on the brand.
Q3. How many runs per question, with repeats?
Why it matters. The same question returns different answers on different runs. One run is an anecdote.
A verifiable answer. A stated number of runs per question and engine, not only a cadence such as weekly.
Q4. Which engines are read, named one by one?
Why it matters. Engines disagree with each other. A result only means something for the engines it names.
A verifiable answer. Each engine named in the service description a buyer reads, not only in a blog post.
Q5. Is the method versioned and dated?
Why it matters. Without a version, a buyer cannot tell whether two reports were measured the same way.
A verifiable answer. A version number and a date on the method, with changes published before they are used.
Measurement and evidence, Q6 to Q10
Q6. Is a baseline measured before the work starts?
Why it matters. Movement can only be shown against a starting point measured the same way.
A verifiable answer. A named baseline deliverable at the start of the engagement, on a fixed question set.
Q7. Can the buyer see the raw answer text behind every result?
Why it matters. A rate is a summary. The answer text is the evidence.
A verifiable answer. Every observed answer handed over as the engine returned it, not only a score.
Q8. Are the sources each engine cited shown per answer?
Why it matters. The cited sources are where the fix usually is.
A verifiable answer. The cited URLs listed against each answer, not only an aggregate citation rate.
Q9. Is variance between runs reported?
Why it matters. If runs disagree, an average hides it.
A verifiable answer. The run to run spread or the disagreement between runs printed beside each rate.
Q10. Is the measurement repeatable by the buyer or a third party?
Why it matters. A measurement nobody else can repeat has to be taken on trust.
A verifiable answer. Enough published method that a third party could rerun it and check the arithmetic.
Deliverables, Q11 to Q15
Q11. What exactly arrives, named as artefacts?
Why it matters. A service category is not a deliverable.
A verifiable answer. A list of named artefacts, for example a baseline report, a fix list and a roadmap.
Q12. When does each artefact arrive?
Why it matters. A timeline in months is not a delivery date.
A verifiable answer. A stated turnaround for a named artefact, for example findings within a set number of hours.
Q13. Who does the work?
Why it matters. A buyer is paying for people as much as for a method.
A verifiable answer. A stated team, or a named lead for the engagement.
Q14. Does the buyer get data access, and keep it?
Why it matters. A dashboard that disappears at cancellation leaves the buyer with nothing to compare against.
A verifiable answer. Access to the data during the work, with retained access or an export stated for after it.
Q15. What does the buyer keep when the engagement ends?
Why it matters. Work product that stays with the vendor is not work product the buyer owns.
A verifiable answer. A written statement of what the buyer keeps on exit.
Commercial terms, Q16 to Q20
Q16. Is an entry price published?
Why it matters. Without a price a buyer cannot compare before a call.
A verifiable answer. The company's own price for a named offer, not an industry survey figure.
Q17. Can the buyer pay without a call?
Why it matters. Every call is a step between the buyer and the evidence.
A verifiable answer. A checkout for at least one offer.
Q18. Is the minimum term published?
Why it matters. A minimum term is the real price of a retainer.
A verifiable answer. The minimum commitment in months or weeks, in writing.
Q19. Is there a written guarantee or refund term?
Why it matters. Answers vary run to run, so the refund is the part of the risk a vendor can carry.
A verifiable answer. The trigger, the window and the exclusions of any refund, in writing.
Q20. How does the buyer cancel?
Why it matters. The notice period decides how long a buyer pays after deciding to stop.
A verifiable answer. A stated notice period or the words month to month, in writing.
Proof, Q21 to Q25
Q21. Are there named B2B software case studies?
Why it matters. Anonymous cases cannot be checked.
A verifiable answer. A named client in B2B software with its own case page.
Q22. Does a case study publish its measurement?
Why it matters. A result without its baseline and method is a claim, not a finding.
A verifiable answer. A dated baseline, the method, the window and the numbers the claim rests on.
Q23. Is original research published with its sample and method?
Why it matters. Research shows whether a company measures anything at all.
A verifiable answer. A study with a stated sample size, collection window and classification rules.
Q24. Is data published openly with a licence or DOI?
Why it matters. Open data can be checked by anyone; a download with no licence cannot be cited safely.
A verifiable answer. A dataset with a licence and a DOI.
Q25. Is there a published correction route for the site's own claims?
Why it matters. Pages go stale. A correction route is how errors get fixed.
A verifiable answer. A stated way to report an error and a statement that corrections are made.
The board
Full Yes answers out of 25, with the Yes count in each group. Sort by any heading. Broadcastwell is pinned first and tagged because it publishes this page.
| BroadcastwellOur listing | 20 | 4 | 5 of 5 | 4 of 5 | 3 of 5 | 4 of 5 | 4 of 5 |
| Citant | 12 | 5 | 2 of 5 | 1 of 5 | 3 of 5 | 4 of 5 | 2 of 5 |
| Optimist | 10 | 2 | 1 of 5 | 1 of 5 | 4 of 5 | 3 of 5 | 1 of 5 |
| WebFX | 7 | 5 | 1 of 5 | 1 of 5 | 2 of 5 | 2 of 5 | 1 of 5 |
| iPullRank | 6 | 2 | 1 of 5 | 1 of 5 | 2 of 5 | 1 of 5 | 1 of 5 |
| Omniscient Digital | 5 | 4 | 1 of 5 | 0 of 5 | 1 of 5 | 1 of 5 | 2 of 5 |
| Siege Media | 5 | 3 | 0 of 5 | 1 of 5 | 1 of 5 | 1 of 5 | 2 of 5 |
| Minuttia | 4 | 9 | 2 of 5 | 0 of 5 | 0 of 5 | 0 of 5 | 2 of 5 |
| Go Fish Digital | 2 | 2 | 0 of 5 | 0 of 5 | 1 of 5 | 0 of 5 | 1 of 5 |
| Single Grain | 2 | 5 | 1 of 5 | 0 of 5 | 1 of 5 | 0 of 5 | 0 of 5 |
| First Page Sage | 0 | 5 | 0 of 5 | 0 of 5 | 0 of 5 | 0 of 5 | 0 of 5 |
By group, across the ten agencies: method 9 Yes of 50, measurement and evidence 5 Yes of 50, deliverables 15 Yes of 50, commercial terms 12 Yes of 50, proof 12 Yes of 50. The market answers what it delivers and what it charges; it rarely answers how a result can be checked.
The questions nobody answers
The nine questions with no Yes from any of the ten agencies.
| Question | Yes | Partial | Broadcastwell |
|---|---|---|---|
| Q1. Is the question set published? | 0 | 6 | Yes |
| Q3. How many runs per question, with repeats? | 0 | 5 | Yes |
| Q5. Is the method versioned and dated? | 0 | 3 | Yes |
| Q7. Can the buyer see the raw answer text behind every result? | 0 | 0 | Yes |
| Q8. Are the sources each engine cited shown per answer? | 0 | 1 | Yes |
| Q9. Is variance between runs reported? | 0 | 1 | Yes |
| Q10. Is the measurement repeatable by the buyer or a third party? | 0 | 1 | Partial |
| Q17. Can the buyer pay without a call? | 0 | 0 | Yes |
| Q24. Is data published openly with a licence or DOI? | 0 | 1 | Yes |
Where Broadcastwell's own site falls short
Broadcastwell answers 20 of 25 on its own public pages. These are the five it does not fully answer, with what the site says instead.
| Question | Verdict | What the site says | Page |
|---|---|---|---|
| Q10. Is the measurement repeatable by the buyer or a third party? | Partial | Published so a reader can reproduce its shape and check the arithmetic; implementation detail is not published | /methodology |
| Q12. When does each artefact arrive? | Partial | Category Audit results within 48 hours of category confirmation; the Diagnostic has no promised delivery time | /terms |
| Q13. Who does the work? | Partial | A team of twelve across six roles is stated; no named lead per engagement | /company-facts |
| Q20. How does the buyer cancel? | Partial | Month to month after the initial term is stated; notice periods sit in the signed agreement | /terms |
| Q22. Does a case study publish its measurement? | No | Client baseline data and outcomes are kept private, so no case publishes its measurement | /vantaa-tax-ai-visibility-case-study |
Every answer, with its source
Each company's Yes and Partial answers, with the page that answers each one. Open a company to see its full list.
Broadcastwell: 20 Yes, 4 Partial
- Yes Q1. Complete question sets are public in the Absence Index release files; each client's own set is agreed before the run and stays private. index.broadcastwell.com/downloads.html
- Yes Q2. Brand matching on word boundaries, domain matching on hostnames, published. broadcastwell.com/methodology
- Yes Q3. Three scheduled runs per question and engine, up to two more where verdicts disagree. broadcastwell.com/methodology
- Yes Q4. ChatGPT, Claude, Perplexity, Google AI Overviews and Google AI Mode. broadcastwell.com/pricing
- Yes Q5. Version 1.1, dated 8 September 2026, with a change log. broadcastwell.com/methodology
- Yes Q6. The program is measured against a baseline on agreed questions, engines and method version. broadcastwell.com/terms
- Yes Q7. Every observed answer delivered as the engine returned it. broadcastwell.com/methodology
- Yes Q8. Every source each answer cited is delivered with it. broadcastwell.com/methodology
- Yes Q9. The run to run spread is reported beside the rate. broadcastwell.com/methodology
- Partial Q10. Published so a reader can reproduce its shape and check the arithmetic; implementation detail is not published. broadcastwell.com/methodology
- Yes Q11. Named artefacts per offer, including a receipts appendix and machine-readable files. broadcastwell.com/pricing
- Partial Q12. Category Audit results within 48 hours of category confirmation; the Diagnostic has no promised delivery time. broadcastwell.com/terms
- Partial Q13. A team of twelve across six roles is stated; no named lead per engagement. broadcastwell.com/company-facts
- Yes Q14. Results in the buyer's account, with export and revocable connections for the buyer's own tools. broadcastwell.com/pricing
- Yes Q15. The question set, citation logs, observed answers and reports belong to the buyer. broadcastwell.com/terms
- Yes Q16. $490 Category Audit. broadcastwell.com/pricing
- Yes Q17. Card checkout for the $490 Category Audit and the $990 Diagnostic. broadcastwell.com/pricing
- Yes Q18. The program has a 90-day initial term, then month to month. broadcastwell.com/terms
- Yes Q19. Audit and Diagnostic refundable in full within 30 days of delivery; the 90-day proof gate on the program. broadcastwell.com/terms
- Partial Q20. Month to month after the initial term is stated; notice periods sit in the signed agreement. broadcastwell.com/terms
- Yes Q21. Vantaa Tax, a named B2B software client, with its own case page. broadcastwell.com/vantaa-tax-ai-visibility-case-study
- Yes Q23. 860 answers across 85 companies, with the collection window and scoring rules. broadcastwell.com/state-of-geo
- Yes Q24. Open access under CC BY 4.0 with a DOI. broadcastwell.com/state-of-geo
- Yes Q25. Corrections to hello@broadcastwell.com, made on the page with the date. broadcastwell.com/company-facts
Read 17 September 2026.
Go Fish Digital: 2 Yes, 2 Partial
- Partial Q4. Names AI Overviews and ChatGPT only; no full engine list. gofishdigital.com/services/generative-engine-optimization/
- Yes Q11. Semantic content audit, AI Overview analysis, technical fixes, digital PR placements. gofishdigital.com/services/generative-engine-optimization/
- Partial Q14. Barracuda platform clients use; retained access after engagement not stated. gofishdigital.com/marketing-intelligence
- Yes Q21. SimpleTexting named, B2B SMS software, organic SQL result. gofishdigital.com/results/
Read 17 September 2026.
iPullRank: 6 Yes, 2 Partial
- Partial Q1. Keyword portfolio maps synthetic query gaps; set itself not published. ipullrank.com/ai-search-strategy-program
- Yes Q4. ChatGPT, Claude, Perplexity and Google AI Overviews named individually. ipullrank.com/ai-search-metrics
- Yes Q6. "AI Search visibility baseline" listed as a first-stage deliverable. ipullrank.com/services/geo
- Yes Q11. Six named artefacts: keyword portfolio, content audit, content plan, measurement plan, audit, roadmap. ipullrank.com/ai-search-strategy-program
- Yes Q13. Named founder and named role leads attached to the programme. ipullrank.com/ai-search-strategy-program
- Partial Q14. Measurement plan sets up a data pipeline and dashboards; ownership not stated. ipullrank.com/ai-search-strategy-program
- Yes Q16. "Starting at $15,000/month". ipullrank.com/ai-search-strategy-program
- Yes Q23. 13.1bn search events, named panel source, 15-month window, classification rules. ipullrank.com/zero-click-behavior-analysis-q3-2026
Read 17 September 2026.
Optimist: 10 Yes, 2 Partial
- Partial Q3. "Weekly AI visibility tracking" is a cadence, not a run count. yesoptimist.com/full-service-seo/
- Yes Q4. ChatGPT, Perplexity, Gemini, Claude, Google AI Overviews named. yesoptimist.com/ai-search-optimization-services/
- Yes Q6. Audit of current visibility across the five models precedes work. yesoptimist.com/ai-search-optimization-services/
- Yes Q11. Articles, refreshes, landing pages, links, technical sprints, schema; 4 to 5 per month. yesoptimist.com/full-service-seo/
- Partial Q12. Results horizon 1 to 3 months; no per-artefact delivery date. yesoptimist.com/full-service-seo/
- Yes Q13. Named founder leads strategy and calls; named human editorial review. yesoptimist.com/full-service-seo/
- Yes Q14. Custom reporting dashboard; clients retain dashboards and data on exit. yesoptimist.com/full-service-seo/
- Yes Q15. On cancellation clients keep roadmap, content, links, schema, dashboards, data. yesoptimist.com/full-service-seo/
- Yes Q16. $7,500 roadmap; $3,000/mo advisory; $4,000/mo full service. yesoptimist.com/ai-search-optimization-services/
- Yes Q18. "Month-to-month with no minimum commitment". yesoptimist.com/full-service-seo/
- Yes Q20. 30-day notice to cancel. yesoptimist.com/full-service-seo/
- Yes Q21. Stampli named, B2B financial automation software. yesoptimist.com/case-study-stampli/
Read 17 September 2026.
Siege Media: 5 Yes, 3 Partial
- Partial Q4. Engines named in a research post, not in the service description. siegemedia.com/research/ai-traffic-conversion-rate
- Yes Q6. "Baseline LLM share-of-voice benchmark" in the Month 1 to 2 phase. siegemedia.com/services/geo
- Yes Q11. Monthly share-of-voice reporting, citation tracking per asset, referral benchmarking, quarterly reviews. siegemedia.com/services/geo
- Partial Q12. Phases dated in months; no artefact delivery dates. siegemedia.com/services/geo
- Partial Q13. Roughly 80 named staff listed; no team size for an engagement. siegemedia.com/about
- Yes Q18. "6+ month minimum commitment". siegemedia.com/services/geo
- Yes Q21. Zendesk, Zapier, Zoom, HubSpot, Airtable named. siegemedia.com/work/all
- Yes Q23. 78 sites, Jan to May 2026, GA4 key-event rate, screening thresholds. siegemedia.com/research/ai-traffic-conversion-rate
Read 17 September 2026.
Omniscient Digital: 5 Yes, 4 Partial
- Partial Q1. 200 BOFU prompts, six intent types, how built; set not published, research not service. beomniscient.com/blog/g2-acquisition-ai-citation-share/
- Partial Q3. Single run, February 2026; no repeat count. beomniscient.com/blog/g2-acquisition-ai-citation-share/
- Yes Q4. ChatGPT, Perplexity, Gemini, AI Mode, AI Overviews named. beomniscient.com/blog/content-types-cited-in-llms/
- Partial Q5. Study dated February 2026; no method version. beomniscient.com/blog/g2-acquisition-ai-citation-share/
- Partial Q6. Opportunity and citation analysis precede work; not called a baseline. beomniscient.com/services/generative-engine-optimization/
- Yes Q11. Opportunity analysis, citation analysis, technical assessment, roadmap, monitoring. beomniscient.com/services/generative-engine-optimization/
- Yes Q16. "Full-service engagements start at $10,000 a month". beomniscient.com/services/generative-engine-optimization/
- Yes Q21. Jasper, Smartling, SuperTokens, SpotDraft, Lokalise, 360Learning named. beomniscient.com/case-studies/
- Yes Q23. Ten own studies listed with prompt counts and citation counts. beomniscient.com/research/
Read 17 September 2026.
Minuttia: 4 Yes, 9 Partial
- Partial Q1. Prompt counts and four prompt types published; set itself not published. study.minuttia.com/methodology
- Yes Q2. Deterministic match to a 5,983-brand corpus; ambiguous names excluded. study.minuttia.com/methodology
- Partial Q3. Single execution per prompt on Google; not stated for other surfaces. study.minuttia.com/methodology
- Yes Q4. Google Search + AIO, ChatGPT, Gemini, Claude, Perplexity named. study.minuttia.com/methodology
- Partial Q5. Wave-1 24 Aug 2026, snapshot 26 Aug 2026; no version number. study.minuttia.com/methodology
- Partial Q9. States results vary by time, location, account and repetition; no interval given. study.minuttia.com/methodology
- Partial Q10. Method described and aggregate data downloadable; no reproduction instructions. study.minuttia.com/methodology
- Partial Q11. Service categories listed; no itemised artefacts. minuttia.com/geo-agency/
- Partial Q13. Managing Director named; no team size for the work. minuttia.com/geo-agency/
- Yes Q21. Meilisearch, Toggl, Vowel, Respona, Viral Loops named. minuttia.com/case-studies/
- Partial Q22. Names tools and date range; conversion and visibility maths undisclosed. minuttia.com/case-studies/meilisearch/
- Yes Q23. AI Visibility Data Study, 24,067 prompts, five surfaces. study.minuttia.com/
- Partial Q24. "Download all aggregate data" offered; no licence and no DOI. study.minuttia.com/
Read 17 September 2026.
First Page Sage: 0 Yes, 5 Partial
- Partial Q4. ChatGPT, Claude, AI Overviews named in one case; not on the service page. firstpagesage.com/fps-geo-case-study-nearshore-software-development-company/
- Partial Q11. Six optimisation elements listed as roles, not as artefacts. firstpagesage.com/generative-engine-optimization-services/
- Partial Q21. Cadence Design Systems named; GEO cases otherwise anonymised. firstpagesage.com/case-studies/
- Partial Q22. Names tracking tool and share over three years; no query set or baseline method. firstpagesage.com/fps-geo-case-study-nearshore-software-development-company/
- Partial Q23. Proprietary campaign data Q1 2021 to Q3 2025; no sample size or collection method. firstpagesage.com/reports/seo-roi-statistics-fc/
Read 17 September 2026.
Single Grain: 2 Yes, 5 Partial
- Partial Q1. "Find Your Best AI Questions & Keywords" describes building a set, not the set. singlegrain.com/agency/generative-engine-optimization-services/
- Yes Q4. ChatGPT, AI Overviews, Gemini, Perplexity, Claude, Copilot named. singlegrain.com/agency/generative-engine-optimization-services/
- Yes Q11. Content briefs, expert guides, Q&A hubs, author bios, schema, GEO dashboards. singlegrain.com/agency/generative-engine-optimization-services/
- Partial Q12. Citations improve in 2 to 3 months, gains in 4 to 6; no artefact dates. singlegrain.com/agency/aeo-services/
- Partial Q13. "Dedicated AEO team" stated; no size and no named owner. singlegrain.com/agency/aeo-services/
- Partial Q14. Custom dashboards provided; retained access not stated. singlegrain.com/agency/aeo-services/
- Partial Q21. SmartRent and Learning A-Z named; B2B software status not stated. singlegrain.com/case-studies/
Read 17 September 2026.
WebFX: 7 Yes, 5 Partial
- Partial Q1. 50 to 2,000 tracked prompts per tier published; set and construction not published. webfx.com/seo/services/ai-search-optimization/
- Partial Q3. Each prompt tested across all ten engines; no repeat run count. webfx.com/seo/services/ai-search-optimization/
- Yes Q4. Ten engines named: ChatGPT, Gemini, Perplexity, Copilot, Claude, Meta AI, Grok, DuckDuckGo, Apple Intelligence, Bing. webfx.com/seo/services/ai-search-optimization/
- Yes Q6. Baseline AI visibility audit, 500 to 50,000 prompts by tier. webfx.com/seo/services/ai-search-optimization/
- Yes Q11. Baseline audit, monthly scoring, brand perception analysis, content pieces per tier. webfx.com/seo/services/ai-search-optimization/
- Partial Q12. Monthly optimisation cadence over six months; no artefact dates. webfx.com/seo/services/ai-search-optimization/
- Yes Q13. Dedicated consultant per plan; "750+ SMEs" stated. webfx.com/seo/pricing/
- Partial Q14. Dashboard with unlimited seats and SSO; access after cancellation not stated. webfx.com/seo/services/ai-search-optimization/
- Yes Q16. Tiers from $3,000 to $68,250+ per month. webfx.com/seo/services/ai-search-optimization/
- Yes Q18. Six-month minimum commitment. webfx.com/seo/services/ai-search-optimization/
- Partial Q21. One client tagged SaaS & Software B2B; the case is paid media, not AI search. webfx.com/case-studies/
- Yes Q23. 2,371,151 keywords, classification by length, intent and vertical. webfx.com/blog/seo/where-and-why-google-ai-overviews-appear-2-3m-keywords-study/
Read 17 September 2026.
Citant: 12 Yes, 5 Partial
- Partial Q1. Ten agreed target queries fixed at onboarding and reused; set not published. citant.ai/services/ai-visibility/
- Yes Q2. Three states: named, retrieved but not named, absent. citant.ai/services/ai-visibility/
- Partial Q3. One run per platform per month; no repeat count per question. citant.ai/services/ai-visibility/
- Yes Q4. ChatGPT, Perplexity, Gemini, Claude, Copilot, AI Overviews named. citant.ai/services/ai-visibility/
- Partial Q5. Method "last verified August 2026"; no version number. citant.ai/services/ai-visibility/
- Yes Q6. Audit sets a baseline Share of Model before the pilot. citant.ai/services/ai-search-audit/
- Partial Q8. Ghost-citation rate per platform where sources surface; not per answer. citant.ai/services/ai-visibility/
- Yes Q11. Baseline Share of Model, 4-Layer scorecard, prioritised fix list. citant.ai/services/ai-search-audit/
- Yes Q12. Delivered within 72 hours of kickoff; nine-week pilot. citant.ai/services/ai-search-audit/
- Yes Q13. Founder named as sole public practitioner on the work. citant.ai/about/
- Partial Q15. Underlying exports available on request; no buyer-side export. citant.ai/case-studies/ai-toolbox/
- Yes Q16. $440 audit, $7,875 nine-week pilot in three payments of $2,625, retainer from $5,000/mo. citant.ai/pricing/
- Yes Q18. Nine-week pilot; three-month retainer minimum. citant.ai/pricing/
- Yes Q19. Named in 3 of 10 queries on 2 of 6 platforms in nine weeks, or full refund. citant.ai/pricing/
- Yes Q20. Retainer releases on 30 days notice after the three-month minimum. citant.ai/pricing/
- Yes Q22. Dated baseline 27 Apr 2026, execution window, five tracked points, dated exports. citant.ai/case-studies/ai-toolbox/
- Yes Q25. States that every number on its site carries a source and invites readers to flag one that does not. citant.ai/about/
Read 17 September 2026.
Cite this page: The GEO agency RFP: 25 questions, and who answers them in public. Broadcastwell. Last reviewed 18 September 2026. https://broadcastwell.com/geo-agency-rfp
What this page does not prove
This page records what each company publishes on its own site on the date beside it. It does not measure the quality of anyone's work, it does not rank outcomes, and a company that publishes less is not necessarily worse at the job. Pages change; a cell is only true for its read date.
An agency that answers a question on a call, in a proposal or in a contract may answer it well; this page only records what a buyer can read before any of that. The ten agencies were chosen by Broadcastwell from the companies on its ranking pages; they are not a random sample of the market.
Broadcastwell's own limitations
- Three named clients.
- One early directional result.
- No third party reviews yet.
- Founded 2026.
- A team of twelve.
Questions
What should I ask a GEO agency before hiring one?
Ask the 25 questions on this page, in five groups: method, measurement and evidence, deliverables, commercial terms and proof. The evidence group matters most, because it decides whether you can check a result yourself: can you see the raw answer text behind every result, are the cited sources shown per answer, is the variance between runs reported, and could a third party repeat the measurement. Across the ten agencies read, those four questions drew not a single Yes answer.
Which GEO agency answers the most of these questions in public?
Citant, with 12 of 25 answered with a Yes on its own site. The median across the ten agencies read is 5 of 25. Broadcastwell scores itself 20 of 25 on the same rule, from its own public pages, and lists the questions it does not fully answer.
Can I pay a GEO agency without a sales call?
Not at any of the ten agencies read here. Every one routes a buyer through a form, a call or an email first. Broadcastwell sells the $490 Category Audit and the $990 Diagnostic by card checkout. The page on buying without a call lists every offer in this market that can be bought that way.
Is a Partial as good as a Yes?
No. A Partial means part of the answer is on the site, for example a cadence of weekly tracking where the question asks how many runs there are per question. The counts on this page use full Yes answers only and print Partials separately.
Can I use these questions in my own RFP?
Yes. Copy all 25 with the button on this page, print the page, or download the CSV, which carries every verdict with its source and read date. Please cite the page if you publish the list.
How did Broadcastwell score itself?
By the same rule, from its own public pages only: /pricing, /terms, /methodology, /company-facts and the case pages. Every Yes carries the URL that answers it. It answers 20 of 25; the five it does not fully answer are listed on this page, including the one it answers No: Q22, does a case study publish its measurement?.
The evidence questions are the ones this market leaves blank. The $490 Category Audit answers them for your own category: every observed answer, its engine, its date and the sources it cited.
See where the engines place you
See which buyer questions leave you out, who is named instead of you, the sources visible in those answers, and the three changes worth testing first. The Category Audit covers one category, ten buyer questions, five engines and three measured runs, with findings within 48 hours of the category being confirmed.
Start with the $490 Category Audit$490 once. Results appear in your account. The full $490 credits against the Diagnostic within 30 days. Refundable in full, no questions asked, within 30 days of delivery of the findings.
Get the Diagnostic, $990$990 once. 35 questions, five engines, three scheduled runs per question and engine pair, 525 scheduled observed answers, adaptive runs extra. The full $990 credits against the program's first monthly invoice.
Every figure comes from the published method, v1.1.