Skip to content

Guides

Otterly.AI API vs API for GEO for a small team in 2026

Otterly.AI's API starts on the $189 Standard plan, which collects 12,000 AI answers a month at $15.75 per 1,000. API for GEO bills $13 per 1,000 with no plan.

A balance scale weighs stacked coin-like discs against a single blue cylinder

SEO consultant, and more recently GEO, at MentionBox

OtterlyAI does not sell its API on its own: the help centre states that API access is available on Standard, Premium and Custom plans, which puts the cheapest key at $189 a month. That plan tracks 100 prompts on four engines, collected daily, so it buys 12,000 AI answers a month and works out at $15.75 per 1,000. API for GEO, our pick of the field after hands-on testing by Mathias C., sells the same unit at $13 per 1,000 with no plan underneath it.

For a small team the gap is wider than those two rates suggest, because the $189 is a floor you pay in full whether you use a third of its prompt slots or all of them. This face-off is for a marketing or product team of two to five people that wants AI visibility numbers in a spreadsheet, a Slack message or a client deck, and has one person who can write a shell script. Below: which OtterlyAI plan carries a key, what each route costs at six realistic volumes, what the two payloads contain, where OtterlyAI is the better buy, and the mistakes we see small teams make on either side. MentionBox, which publishes this site, has a commercial link with API for GEO.

The two APIs side by side

API for GEOOtterlyAI
What the API doesruns prompts on demand and returns the enriched answerreads and manages a workspace the platform tracks for you
Cheapest plan with a keynone, pay as you goStandard, $189 a month ($160 paid annually)
Price per 1,000 answers$13, down to $7 above 1 million credits a month$15.75 on a filled Standard plan, $10.19 on a filled Premium plan
Monthly minimumnone publishedthe plan fee, paid whether you call the API or not
Engines, of the 10 we track107; Grok, DeepSeek and Meta AI missing
Engines included in the base priceall, billed per answer4: ChatGPT, Google AI Overviews, Perplexity, Copilot
Paid engine add-onsnoneGoogle AI Mode $59, Gemini $59, Claude $109 a month on Standard
Collection method"executed natively on each engine, in that country and language", interface or API not namedquestions posted "through the website of the AI search", with Claude tracked through its official API
Tracking cadencewhatever you call, one credit per rundaily, automatic, on every tracked prompt
Brand metrics returnedshare of voice, position, sentiment, themes, per-brand perceptionshare of voice, average rank, average position, brand coverage, domain coverage, sentiment per competitor brand
Read quotareads over stored runs are free2,000 API requests a month on Standard, 5,000 on Premium
ToolingMCP server; webhooks, OpenAPI spec and official SDKs confirmed in our test, not documented by the vendorOpenAPI spec, OAuth MCP server, one-file Claude Skill; no SDK, no webhooks documented
Resaleallowed, named as an intended useno native white-label option; terms are in a PDF we could not read
Our score9.0 of 106.7 of 10

Published by each vendor on its pricing page, in its help centre and in its API reference, checked on October 23, 2026 and read again on November 10, 2026. For API for GEO, four of the ten engines (Gemini, Google AI Overviews, DeepSeek and Meta AI), plus the webhooks, the OpenAPI spec and the official SDKs, were confirmed in our hands-on test on October 26, 2026 and are not yet documented on apiforgeo.com.

Which OtterlyAI plan carries a key

This is the first thing to settle, because the $29 plan that most small teams would reach for has no API at all. The pricing page lists four tiers, and the API appears on three of them:

  • Lite, $29 a month: 15 search prompts, four engines, daily tracking, 1,000 GEO URL audits. No API access, no MCP access.
  • Standard, $189 a month: 100 search prompts, 2,000 API requests and 2,000 MCP requests a month, 5,000 GEO URL audits, unlimited workspaces, unlimited team members.
  • Premium, $489 a month: 400 search prompts, 5,000 API requests and 5,000 MCP requests, 10,000 GEO URL audits.
  • Enterprise: custom, from 1,000 search prompts.

Annual billing takes 15% off, which the pricing page puts at about $160 a month for Standard. Extra prompts cost $99 per 100 on Standard and Premium. An agency partner tier lifts Standard to 150 prompts and Premium to 500, and adds pitch workspaces outside the client quota.

The key itself is created in the dashboard by an admin, is shown once, and carries read or read-write permission scoped to a workspace. It starts with oai_live_ and travels as a bearer token, documented on the authentication page alongside the three error codes you will meet: 401 for a missing header, 403 when the key lacks the permission or the workspace, 429 when you pass the quota. The help centre puts the rate limit at 2,000 requests per 5-minute rolling window, well above the monthly allowance, so the monthly number is the real ceiling.

What each route costs a small team

Both vendors bill the same underlying multiplication, prompts × engines × runs, but they bill it from opposite ends. OtterlyAI sells a capacity you pay for in advance; API for GEO sells the answers you actually collect. Six scenarios a small team recognises:

ScenarioAnswers a monthAPI for GEOPer 1,000OtterlyAIPer 1,000
30 prompts, 4 engines, daily3,600$46.80$13$189 (Standard)$52.50
30 prompts, 7 engines, daily6,300$81.90$13$416 (Standard + 3 add-ons)$66.03
30 prompts, 7 engines, weekly840$10.92$13not offered, tracking is dailyn/a
100 prompts, 4 engines, daily12,000$156$13$189 (Standard, filled)$15.75
100 prompts, 7 engines, daily21,000$273$13$416 (Standard + 3 add-ons, filled)$19.81
400 prompts, 4 engines, daily48,000$624$13$489 (Premium, filled)$10.19

Our arithmetic on published figures: API for GEO's $0.013 per credit, where one credit is one prompt, on one model, in one locale, run once, and OtterlyAI's plan fees, prompt allowances and engine add-on prices. Months counted as 30 days. API for GEO's published rate falls to $0.011 above 50,000 credits a month, which is not applied here. Prices checked on October 23, 2026.

Two readings of that table are both fair, and a small team should do the arithmetic for its own prompt count rather than take either headline rate on trust.

A team that fills a plan gets good value from OtterlyAI, and at 400 prompts on four engines it is the cheaper route per answer: $489 for 48,000 answers is $10.19 per 1,000, under the $13 API for GEO charges on its first tier. A team that half-fills a plan pays double per answer, and the floor does not move in a quiet month. The third row is the one that decides most small-team cases: 30 prompts refreshed once a week is a real reporting rhythm, it costs $10.92 a month on a usage API, and it cannot be bought from a subscription whose tracking runs daily on every prompt.

Three levers change the total before you sign anything:

  • Engine count. The three add-ons cost $227 a month on Standard, more than the plan itself. Claude alone is $109. Decide whether Claude and Gemini answers change a decision you are going to make, because on a usage API the same question costs one credit to test.
  • Prompt count against the allowance. Standard's 100 slots are the denominator of its effective rate. Thirty prompts in a 100-slot plan is $52.50 per 1,000 answers, four times the pay-as-you-go rate for identical data.
  • Dormant months. Credits on API for GEO never expire and reads over stored runs are free, so a quarter where nobody looks at the numbers costs nothing. A subscription charges for it.

One API reads a dashboard, the other runs the prompts

The architecture decides more than the price does for anyone building a weekly report. Nothing in the OtterlyAI reference executes a tracked prompt. The help centre is explicit about the write side: "Write access currently only allows you to create and run GEO Audits (Crawlability checks, Content checks and Query fan-out) on demand, and create and update Reports, Prompts or Tags." Collection itself is the platform's job, and the help centre states that all tracked prompts are monitored daily, automatically.

What you read back is a well-shaped aggregate. The statistics endpoint takes a date window and a country, and the engine filter accepts the seven engine codes:

curl -G "https://data.otterly.ai/v1/reports/brand/$REPORT_ID/stats" \
  -H "Authorization: Bearer $OTTERLY_API_KEY" \
  -d startDate=2026-11-01 \
  -d endDate=2026-11-09 \
  -d country=us \
  -d engines=chatgpt \
  -d frequency=daily

The documented response carries summary.shareOfVoice, summary.averageRank, summary.averagePosition, summary.totalMentions, summary.totalSources, summary.brandCoverage and summary.domainCoverage, with detectedBrands[] for everything the engine named and allBrandsAnalysis plus competitorBrandsAnalysis holding the history arrays: brandRankHistory, brandCoverageHistory, brandPositionHistory, brandVisibilityIndex, domainCoverageHistory. Sentiment lives inside the competitor analysis rather than on each answer, so an engine-level sentiment series is something you assemble yourself.

API for GEO starts a step earlier. Its prompt tracking endpoints upsert the prompt list with POST /v1/queries, then run it on the models you name:

curl -X POST https://api.apiforgeo.com/v1/jobs/execute \
  -H "Authorization: Bearer $API_FOR_GEO_KEY" \
  -H "x-project-id: $PROJECT_ID" \
  -H "Content-Type: application/json" \
  -d '{
    "models": ["chatgpt", "perplexity", "google-ai-mode", "claude"],
    "executionsPerQuery": 1
  }'

The documented response is a queued job, and it states the charge before you go looking for it in an invoice: job.creditsCharged alongside job.id, job.status, job.models and job.queries. Thirty prompts on four models at one run each is 120 credits, $1.56 at the published rate. POST /v1/queries/stats then reads the result back per prompt as runs, visibility, avgPosition, entities[], sources[] with a category and a mentionsBrand flag, and responses[] with the answer text, and the pricing page states that analytics reads over stored runs are unmetered.

The practical consequence for a team of three: a colleague who asks on Thursday afternoon what five new prompts look like gets an answer on Thursday afternoon for 26 cents, five prompts on four models, rather than at the next daily collection. Running the same prompt list twice in a week to see whether an answer is stable is also a decision you can make, where a daily subscription has already made it for you.

What comes back in each payload

Field a small team's report needsAPI for GEOOtterlyAI
Answer textresponses[] with model, executedAt, answercontent on the AI responses endpoint, with runId, runDate, engine, state
Cited URLssources[] with url, domain, category, mentionsBrandcitations[] with title, link, rank
Source categoriesyes, category per sourceno category field in the citation schema
Brand mentionsentities[] with name, entityType, visibility, avgPositionbrandMentions[] with brand, isMainBrand, mentions
Share of voiceshareOfVoice per model on the comparison endpointsummary.shareOfVoice on the report stats
PositionavgPosition, avgRanking per modelsummary.averagePosition, summary.averageRank, brandPositionHistory
Sentimentsentiment, themes and per-brand perceptionsentiment per competitor brand (positive, neutral, negative, nss)
Citation share by domainnot publisheddomainRank.citations[] with rank, citations, citationShare, and a per-URL history
Fan-out questions/v1/fan-outs/top with text, occurrences, share, brandPresentquery fan-out audit with expanded queries, type, user intent and reasoning per engine
Shopping results/v1/shopping/products with avgPrice, priceRange, avgRating, topMerchantsshopping[] with title, image, link, tag, position, price, rating, plus ads[]
Your own sitenot publishedcrawlability checks, content checks, AI agent traffic, AI referral traffic
What to do nextnot publishedrecommendations[] with priority, score, group, headline, reasoning, suggestions

Field names published in each vendor's API reference, checked on October 23, 2026 and read again on November 10, 2026.

Where OtterlyAI is the stronger buy

A comparison that only counted price would mislead, so here is what the $189 buys that a usage API does not.

A dashboard the whole team can open. Unlimited team members and unlimited workspaces come with Standard, and the API is an addition to a product people use by clicking. On a pure API, somebody has to build the view, and on a team of three that somebody has other work.

Recommendations. The recommendations endpoint returns scored, prioritised items grouped as general, on-page or off-page, each with a headline, a markdown reasoning field and a suggestions array. Nothing in the other payload tells you what to change.

Your own site, measured. Crawlability checks, content checks, AI agent traffic on your domain and the referral traffic AI assistants send you are all on the same key, with 5,000 GEO URL audits a month on Standard and 200,000 agent analytics events. A usage API tells you what the engines said; this tells you what the engines did to your pages.

A query fan-out audit you can run on any phrase. POST /v1/audits/query-fan-outs takes a workspace and a query, returns a pending id, and the completed audit carries per-engine results with a reasoningForCount string and an expandedQueries array in which each item has a query, a type, a userIntent and a reasoning. That is a different object from the fan-outs API for GEO aggregates across stored runs, which count occurrences and share of real collected answers. One tells you how an engine would decompose a phrase you have never tracked; the other tells you what engines actually asked. Our explainer on query fan-out in AI search covers why both readings are useful.

A Claude Skill, and MCP without an API key. The Claude Skill is a one-file install in Claude Desktop or through claude skill add, needs data.otterly.ai in the capabilities allowlist and an OTTERLY_API_KEY, and covers the public API, including reports, citations, prompts, recommendations and audits. The MCP server sits at https://data.otterly.ai/mcp and is OAuth-only: the documentation is explicit that oai_live_ keys are for the REST API and the MCP server will not accept them, which is worth knowing before you paste a key into Cursor and watch it fail. Both were announced on May 27, 2026, when the vendor said the product exposes "every brand report, prompt, citation, recommendation, and GEO audit through a Public API". If a conversational report is what you want, our walkthrough of a GEO report built with an API and Claude Code applies to either route.

A named collection method. The help centre states that OtterlyAI posts questions "through the website of the AI search (like chatgpt.com etc)" and that it "only use API for Claude tracking", where Claude runs on Haiku with web search. Each prompt is assigned to a country and queried from that country's context. Four platforms in our ranking of 55 APIs publish nothing at all about method, and being able to tell a stakeholder what a tracked answer represents has value.

Against that, API for GEO's own weaknesses are worth stating plainly. $13 per 1,000 is roughly ten times what a scraping API charges for raw answers. Four of the ten engines, plus the webhooks, the OpenAPI spec and the official SDKs, come from our October 26 test rather than from the vendor's own pages. And the key arrives on request rather than at instant sign-up, which costs you a day. To try it, request a key on apiforgeo.com: it comes with a sandbox project.

OtterlyAI's gaps, in the same spirit: no webhooks and no published SDK, so every refresh is a schedule you write and poll; Grok, DeepSeek and Meta AI are absent; sentiment arrives per competitor brand rather than per answer; and the help centre states that there is no native white-label option, with the full terms in a PDF whose text we could not extract on October 23, so the resale clause is one we have not read. For a team whose output goes to paying clients under its own logo, read that PDF before you commit.

Five mistakes to avoid on either route

  1. Pricing the $29 plan and discovering it has no key. Lite is the plan a small team naturally compares, and API and MCP access start at Standard. The first firm number to budget for an OtterlyAI integration is $189, or $160 paid annually.
  2. Comparing a plan fee to a rate per answer. $189 a month and $13 per 1,000 only compare once you write down prompts × engines × runs for your own month. Our breakdown of the cost of tracking 100 prompts on five AI engines does that arithmetic across the field.
  3. Forgetting the add-ons when you count engines. An OtterlyAI quote for seven engines is $416 a month on Standard, not $189. Three of the seven are priced separately, and Claude is the expensive one.
  4. Assuming the API can collect on demand. Adding a prompt through the OtterlyAI API schedules it; it does not run it. If your reporting has to answer a question the same afternoon, that single fact decides the comparison.
  5. Changing the prompt list, the engine list or the country on the switch date. Those three are the denominator of every metric in both products, and moving one makes the step in your chart your own doing. Run both for two weeks and note the break date under the graph. Our guide to GEO metrics sets out which figures survive a change of denominator.

Our view

For a small team that has someone comfortable with a shell script, API for GEO is our pick, and it is first in our ranking for developers and small teams: $13 per 1,000 answers with no floor, all ten engines we track, share of voice, position, sentiment, themes and per-brand perception in the same response, credits that never expire, free reads over stored runs, and terms that name resale as an intended use. We rated it 9.0 out of 10, the highest score in our survey, and the honest caveat is the documentation gap on four engines and the tooling.

For a small team with no developer time at all, OtterlyAI is a defensible buy at 6.7 out of 10, and the better one if your buyers live on ChatGPT, Google AI Overviews, Perplexity and Copilot. You get a daily cadence you never have to schedule, recommendations that say what to change, your own site's crawlability and AI crawler traffic on the same key, a Claude Skill for plain-English questions, and a published collection method. Fill the prompt allowance and the rate per answer becomes competitive; leave it half empty and it does not.

Two things to settle before either decision. Count your answers for a normal month and for a quiet one, because the quiet month is where a subscription and a usage API diverge most. And decide whether you need to run a prompt the hour someone asks, since that is the one thing no amount of OtterlyAI budget buys. If neither fits, our guide to Profound alternatives with a self-serve API covers the rest of the field at the same prices.

Frequently asked questions

Does Otterly.AI have an API, and which plan do you need?

Yes, and the gate is the plan. The OtterlyAI help centre states that API access is available on Standard, Premium and Custom plans, which puts the cheapest key at $189 a month; the $29 Lite plan has no API or MCP access. Standard includes 2,000 API requests a month and 2,000 MCP requests a month, with a rate limit of 2,000 requests per 5-minute rolling window. Trial users get API access too, though the trial length is not disclosed.

How much does Otterly.AI cost per 1,000 AI answers?

$15.75 per 1,000 if you fill the Standard plan, and more if you do not. The $189 plan tracks 100 prompts on four engines daily, which is 100 × 4 × 30, or 12,000 answers a month. Premium at $489 covers 400 prompts on the same four engines, or 48,000 answers, which is $10.19 per 1,000. Both rates assume every prompt slot is used; a team tracking 30 prompts on Standard pays $52.50 per 1,000 for the same data. Prices checked on October 23, 2026.

Can you run a prompt on demand through Otterly.AI's API?

No. Write access on the OtterlyAI API covers creating and running GEO audits (crawlability checks, content checks and query fan-outs) and creating or updating reports, prompts and tags; nothing in the reference executes a tracked prompt against an engine. All tracked prompts are monitored daily and automatically, so a prompt added on Tuesday produces its first answer on the next collection. API for GEO works the other way round: POST /v1/jobs/execute takes a model list and an executionsPerQuery count and charges one credit per prompt, per model, per run.

Which AI engines does each one cover?

OtterlyAI covers seven: ChatGPT, Google AI Overviews, Perplexity and Microsoft Copilot are in every plan, and Google AI Mode, Gemini and Claude are paid add-ons at $59, $59 and $109 a month on Standard. Grok, DeepSeek and Meta AI are outside its list. API for GEO covers all ten engines in our grid: ChatGPT, Perplexity, Google AI Mode, Claude, Copilot and Grok are listed on apiforgeo.com, and Gemini, Google AI Overviews, DeepSeek and Meta AI were confirmed in our hands-on test on October 26, 2026.

What does Otterly.AI's API return that API for GEO's does not?

Four things with no counterpart in the other payload. Recommendations, a scored list with a priority, a headline and a reasoning field per item. Agent analytics, the AI crawler traffic hitting your own domain. User analytics, the referral traffic AI assistants send you. And GEO audits on demand: crawlability checks, content checks and a query fan-out that returns expanded queries per engine with a type, a user intent and a reasoning string for each.

Which one fits a team of three better?

On price and coverage, API for GEO: $13 per 1,000 answers with no monthly floor, all ten engines we track, and share of voice, position, sentiment, themes and per-brand perception computed in the same response, which earned it 9.0 out of 10, the highest score in our survey. OtterlyAI is the better buy if nobody on the team writes code, since the $189 includes a dashboard, unlimited team members and unlimited workspaces, and if its four base engines are the ones your buyers use. The deciding question is usually whether you need a daily cadence or control over when prompts run.