Skip to content

Guides

Semrush and Ahrefs AI visibility APIs: what you can pull in 2026

Ahrefs exposes Brand Radar in API v3 at $20 per 1,000 custom prompt checks. Semrush put its AI Optimization data behind an Enterprise-only API on November 4, 2026.

SEO consultant, and more recently GEO, at MentionBox

Of the two SEO suites, only Ahrefs sells an AI visibility API you can buy yourself: the Brand Radar endpoints live inside API v3, open on the Lite plan at $129 a month, and return the answer text, the links it cited and a sentiment object per tracked brand, with custom prompt answers costing $20 per 1,000 through a $50 pack of 2,500 checks. Semrush took longer and chose a different door. Its AI Optimization data reached an API on November 4, 2026, in an Enterprise API available only to Semrush Enterprise subscriptions, while the self-serve Semrush API still returns nothing more than AI Overview presence flags against keywords. API for GEO is our editorial pick when the deliverable has to be a metric per brand that you can resell, after hands-on testing by Mathias C.

This guide is written for the agency or in-house team that already pays one of these two suites and wants to know whether the AI visibility screens it uses can be read by code. Below: the four products hiding behind two brand names, what one AI answer costs on each, the request you actually write, the fields that come back, the engines each key reaches, the refresh rates, and the two contract clauses that decide where the numbers can go. MentionBox, which publishes this site, has a commercial link with API for GEO.

The four products behind the two names

Most comparisons of these two brands put the dashboards side by side. The API question splits each brand in half, because the screens and the endpoints are not sold on the same plan.

ProductWhat its API returnsWhose promptsEntry ticket
Ahrefs Brand Radar API, custom promptsanswers to prompts you configure, with links and sentimentyoursLite at $129 a month, 150 checks included
Ahrefs Brand Radar API, Ahrefs indexanswers to People Also Ask questions from the Ahrefs keyword indexAhrefs'Brand Radar AI add-on from $199 a month
Semrush Enterprise APIprompt tracking from AI Optimization projectsyoursSemrush Enterprise, price on quote
Semrush API, SEO and ProjectsAI Overview and Bing AI presence flags per keywordneither, these are keywordsSEO Business or Semrush One Advanced plus an API unit package

Published by each vendor on its pricing page, in its help centre and in its API reference, checked on October 23, 2026, except the Semrush Enterprise API, released on November 4, 2026 and read on November 22, 2026.

The distinction that trips people up is the last line. Semrush sells an AI Visibility Toolkit from a low monthly price, and it is a good product, but its knowledge base describes exports, scheduled reports and dashboards rather than endpoints. The AI data that reaches an API is the Enterprise AIO data, and the Enterprise plan is quoted rather than listed.

On the Ahrefs side the split is between two sources of prompts rather than two plans. Custom prompts are questions you write, billed in checks, and the answers are free to read by API. The Ahrefs index is a database its help centre describes as "405+ million search-backed prompts across 7 AI platforms", built from People Also Ask questions, billed in API units, and gated behind the Brand Radar AI add-on.

The APIs side by side

API for GEOAhrefs Brand RadarSemrush Enterprise APISemrush API
What the key buysone prompt run on one engine, enrichedchecks, or reads over the Ahrefs indexreads over your AIO projectreads over the Semrush keyword database
Price per 1,000 AI answers$13, down to $7 at volume$20 on the Basic check packon quoteno AI answers returned
Entry ticketpay as you go, no minimum published$129 a month, LiteSemrush Enterprise, on quoteBusiness plan plus a unit package
Engines, of the 10 we track10891, as a flag
Answer textyesresponseGet Raw Responsesno
Share of voiceyes/sov-overview, /sov-historyGet Brand Visibilityno
Sentimentyesbrand_sentiment per mentionGet Brand Sentimentno
Fan-out questionsown endpointssearch_queriesnot publishedno
Self-serve signupyesyesnoyes, API as an add-on
Webhooksyesnot publishednot publishednot published
SDKsofficial SDKsnone publishednone publishednone published
OpenAPI specyesdocs.ahrefs.com/openapi.jsonnot publishednot published
MCP serveryesyes, with tool groupsAIO data not listedyes, read-only
Rate limit300 requests a minute60 requests a minute2 a second, 150 an hour10 a second
Resaleallowedforbidden without written consentforbidden without written consentforbidden without written consent
Our score9.0 of 106.0 of 103.2 of 101.4 of 10

Published by each vendor on its pricing page and in its API reference, checked on October 23, 2026, with the Semrush Enterprise API read on November 22, 2026. Our scores date from the October grid and so predate that release. 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.

Two rows decide most cases. The self-serve row explains why Ahrefs turns up in integration projects and Semrush does not, since one key arrives after a card payment and the other after a sales cycle. The resale row applies to both suites equally and rules them out of anything with a client login on it, which our guide to a white label AI visibility API covers in detail.

1. Ahrefs Brand Radar API

What it is. Around twenty endpoints inside Ahrefs API v3, grouped into AI visibility (ai-responses, topics, cited-pages, cited-domains), web visibility, overview totals (impressions-overview, citations-overview, mentions-overview, sov-overview) and the matching history calls. Every GET has a POST twin that takes a JSON body for complex filters. The reference is public, the full spec sits at docs.ahrefs.com/openapi.json, and the same reports are reachable through the Ahrefs MCP server at api.ahrefs.com/mcp/mcp, where the tools query parameter selects the brand-radar-simple or brand-radar-entities groups.

Price for 1,000 answers. $20 on the custom prompt route, using the Basic pack of $50 for 2,500 checks. The help centre defines a check as "1 prompt execution × 1 LLM × 1 location" and lists three packs plus overage rates.

Ahrefs check packMonthly priceChecksCost per 1,000 answersOverage per check
Included with Lite$129 (the plan)150n/an/a
Basic$502,500$20.00$0.020
Growth$1007,000$14.29$0.015
Scale$25025,000$10.00$0.010

Pack prices and check allowances published in the Ahrefs help centre and on its pricing page, checked on October 23, 2026. The cost per 1,000 is our division of the pack price by its checks.

Reading those answers back costs nothing, which is the single best commercial detail in either suite: "Pulling prompt data for custom queries by API are free, and does not consume any API units on your subscription plan." The endpoint documentation says the same thing from the other side, that "requests returning only custom prompt data are free, while requests including Ahrefs prompt data follow standard API unit pricing".

The Ahrefs index route prices differently and cannot be reduced to a dollar figure per answer, because Ahrefs publishes API units but not the dollar price of a unit. What it does publish is a cost formula, max(base_cost, per_row_cost * num_rows) with a base of 50 units, a default of 1 unit per field, and 10 units each for response, links and volume. So a select of question, response, links and brand_sentiment costs 22 units a row.

Ahrefs planMonthly API unitsRows per requestRows a month at 22 units
Lite, $129200,000100about 9,090
Standard, $249800,000250about 36,360
Advanced, $4492,000,000500about 90,900

Unit allowances and row caps published in the Ahrefs API v3 help page and plan prices on its pricing page, checked on October 23, 2026. The last column is our division of the allowance by a 22-unit row.

Engines covered. Eight of the ten we track. The data_source parameter accepts chatgpt, gemini, perplexity, copilot, grok, google_ai_overviews, google_ai_mode and claude, plus google_ai_overviews_keywords and google_ai_mode_keywords variants that derive visibility from Google queries instead of prompts. Two engines carry caveats worth reading aloud in a scoping call: Claude is available for custom prompts only, tracked "via API with Sonnet 4.6", and each Claude update consumes 8 checks, which puts a Claude answer at $0.16 on the Basic rate. On Grok, the help centre states that "following recent policy changes by Grok, we are temporarily unable to gather new data".

The request. A bearer token in the header, a comma-separated select, and a data_source that is required rather than optional:

curl -G https://api.ahrefs.com/v3/brand-radar/ai-responses \
  -H "Authorization: Bearer $AHREFS_API_KEY" \
  -H "Accept: application/json" \
  -d select=question,response,links,brand_sentiment,search_queries \
  -d data_source=chatgpt,perplexity \
  -d prompts=custom \
  -d country=us \
  -d date=2026-11-20 \
  -d limit=100

What comes back. The AI responses endpoint documents question, response described as "the model's answer text", links as objects carrying a url and a title, brand_sentiment giving each tracked brand mention a sentiment, an aspect and a snippet, search_queries holding the searches the chatbot ran, volume for the prompt, and since October 14, 2026 a funnel_stage and an intent field. The sentiment values are positive, neutral, negative or mixed. Two gaps are worth naming: search_queries fills only when the data source includes ChatGPT or Perplexity, and no field in the documented response gives a rank or position for your brand inside the answer.

Limits. 60 requests a minute by default, HTTP 429 beyond. No webhooks and no official SDKs appear in the documentation, so scheduling is yours to build, a job our guide to alerts with webhooks and scheduled jobs breaks down. The Web Pages endpoint is Enterprise only. Resale is forbidden. Ahrefs Pte. Ltd. is the contracting party, in Singapore, with a separate US entity for American customers.

For whom. The team that already pays for Ahrefs and wants its AI numbers in a warehouse without a new vendor. At $129 plus $50 a month you get a documented, self-serve, OpenAPI-described key with free reads on your own prompts, which is more than most GEO platforms at four times the price offer.

2. Semrush Enterprise AIO and the API it finally got

What it is. Semrush Enterprise AI Optimization is the suite's AI visibility product, built on a database the product page describes as "317M+ AI prompts and 2.43B+ observed AI responses covering 63M distinct brands and domains". For most of 2026 the page promised to "link AIO data and tailor reports across systems via API" while the developer documentation contained no AIO method at all. That changed on November 4, 2026, when the release notes announced that "the Enterprise API is now available. It gives you programmatic access to your workspaces and AI Optimization (AIO) projects, including prompt tracking data: brand visibility, mentions, sentiment, citations, sources, and AI responses."

Price for 1,000 answers. Not calculable, since Enterprise is quoted and the AIO page ends on a demo booking. The one published price fact is unusually good: the Enterprise API "currently doesn't consume API units", so reads over your own project cost nothing on top of the contract.

Engines covered. Nine of the ten we track, and more besides. The ai_platform parameter on the prompt-tracking endpoints accepts fifteen values: CHATGPT, CHATGPT_PAID, CHATGPT_NO_SEARCH, GOOGLE_AI_OVERVIEW, GOOGLE_AI_MODE, GEMINI, MICROSOFT_COPILOT, CLAUDE, CLAUDE_SEARCH, GROK, DEEPSEEK, PERPLEXITY, NAVER, OPENEVIDENCE and BAIDU_AI_OVERVIEW. Three of those exist nowhere else in our grid of 55 APIs: Naver for Korea, Baidu's AI Overview for China, and OpenEvidence for medical questions. The split between CHATGPT, CHATGPT_PAID and CHATGPT_NO_SEARCH is also a real measurement distinction, since a logged-in paid account and a model answering without web search do not return the same brands. Our explainer on why the same prompt gives different answers covers why that separation changes a trend line.

The request. A v4 key in an Apikey header, and a path that walks from workspace to project before it reaches any data:

curl -G "https://api.semrush.com/apis/v4/enterprise/v1/workspaces/$WS/aio/projects/$PROJECT/prompt-tracking/brand-visibility" \
  -H "Authorization: Apikey $SEMRUSH_API_KEY" \
  -d date_from=2026-11-01 \
  -d date_to=2026-11-21 \
  -d ai_platform=CHATGPT \
  -d limit=1000 \
  -d format=json

The overview page sets out the order: call Get Workspaces for a workspace id, Get AIO Projects for a project id, then check that aio_data_ready returns true before any prompt-tracking call. date_from and date_to are required, ai_platform is optional and "if omitted, results cover all tracked platforms", limit defaults to 100 and tops out at 10,000, the sum of offset and limit must not exceed 50,000, and format takes json or csv. Raw answers come from a sibling path, /prompt-tracking/raw-responses.

Limits. The rate limiting page publishes three tiers: 2 requests a second and 100 an hour per user on workspaces, 5 a second on projects, and 2 a second with 150 an hour on prompt tracking. That hourly ceiling is the number to design around, since 150 calls an hour at 10,000 rows each is plenty for reporting and nowhere near enough for a full history backfill in an afternoon. No webhooks, no SDKs and no OpenAPI file are published, and the Semrush MCP server at mcp.semrush.com/v2/mcp lists the SEO, Trends and Projects APIs without mentioning AIO data.

For whom. The brand team that already has an Enterprise contract and wants its AIO screens in Looker or Snowflake. For anyone else, the entry cost of this API is a sales conversation, and our ranking of AI visibility APIs for agencies weights self-serve access heavily for that reason.

3. The self-serve Semrush API

What it is. The API most people mean when they say "the Semrush API" is the SEO, Projects, Traffic and Local family at developer.semrush.com, sold as an add-on to an SEO Business or Semrush One Advanced subscription plus a package of API units. It is a large and well-documented keyword and backlink API. On AI visibility it returns flags, not answers.

What the AI data actually is. Two things. In Domain Analytics, the analytics tutorial documents that "FK52 shows the total number of the AI Overview SERP feature triggered by the keywords that your domain ranks for" and "FP52 shows the total number of the AI Overview SERP feature that your domain actually ranks in", with the matching Fk and Fp filters on value 52 in the organic keywords report. In Position Tracking, serp_feature_filter accepts a longer list of AI codes.

Semrush SERP feature codeWhat the documentation calls it
aio"A summarized AI-generated overview"
asr"A block that contains links used by Google AI to generate the answer"
rsp"A block that contains a ChatGPT answer"
ctt"A block that contains links referenced by ChatGPT in its quotes"
drp"A block that contains two ChatGPT answers for selection"
aai"An AI-generated response"
aic"An AI-generated chat response"
aim"A brief AI-generated summary"
ais"AI-generated stories"

Codes and descriptions published in the Semrush Position Tracking API reference, read on November 22, 2026.

Price. Not expressible per answer, because no answer comes back. The units are published per report instead: filtering the organic positions or organic visibility reports by SERP feature "requires 100 API units per request", the position tracking management methods also cost 100 units per request, and the organic keywords report costs 10 units a line live or 50 historical. The dollar price of a unit is not published, and units come in recurring packages of 2, 5, 10 or 20 million that renew with the subscription, with unused units lost. Rate limits are 10 requests a second and 10 simultaneous requests per account.

The clause to read before you build. The API usage restrictions state that "you can't cache the information received via API for more than one month without Semrush's express written consent". Any plan to accumulate a multi-year AI visibility history in your own warehouse from this key needs that consent in writing first.

For whom. Teams that want to know which of their keywords trigger an AI Overview and whether their domain appears in it, at keyword scale, inside a suite they already pay for. For prompt-level AI answers, this key is the wrong door, and the AI Mode and AI Overviews APIs we have priced elsewhere do that job from $1.20 per 1,000.

What a real month costs

Take one basket and run it through all four. Ten clients, 30 prompts each, two engines, collected weekly, which is 2,400 AI answers a month.

A month for ten clients, 2,400 answersWhat you getBill
API for GEOanswers plus share of voice, position, sentiment, ranked sources$31.20
Ahrefs Brand Radar, Lite plus Basic packanswers, links and per-mention sentiment, free to read by API$179.00
Semrush Enterprise APIthe same answers plus visibility, mentions, sentiment and citationson quote
Semrush APIno AI answers, AI Overview flags on keywords insteadnot comparable

Our arithmetic on the rates each vendor published, checked on October 23, 2026, under one stated basket. The Ahrefs line is the $129 Lite plan, whose 150 included checks leave 2,250 to buy, covered by the $50 Basic pack of 2,500 checks.

Three things fall out of that table. The Ahrefs bill is dominated by the subscription rather than the data, so the second client costs almost nothing and the first costs $179; at 25,000 checks on the Scale pack the same plan plus $250 brings the all-in rate to about $15 per 1,000. The gap to API for GEO is a factor of about six on this basket and closes as volume rises. And Semrush cannot be budgeted from public information at all, which is a finding rather than an omission.

Which engines each key reaches

Engine, of the 10 we trackAPI for GEOAhrefs Brand RadarSemrush Enterprise APISemrush API
ChatGPTyeschatgptCHATGPT, CHATGPT_PAID, CHATGPT_NO_SEARCHflags only
PerplexityyesperplexityPERPLEXITYno
Geminiyes, from our testgeminiGEMINIno
Google AI Overviewsyes, from our testgoogle_ai_overviewsGOOGLE_AI_OVERVIEWaio flag
Google AI Modeyesgoogle_ai_modeGOOGLE_AI_MODEno
CopilotyescopilotMICROSOFT_COPILOTno
Claudeyesclaude, custom prompts only, 8 checksCLAUDE, CLAUDE_SEARCHno
Grokyesgrok, collection pausedGROKno
DeepSeekyes, from our testnoDEEPSEEKno
Meta AIyes, from our testnonono
Total10891, as a flag
Beyond our gridnonenoneNaver, OpenEvidence, Baidu AI OverviewBing Ask AI, AI chat, AI stories

Engine lists published in each vendor's API reference and help centre, read on November 22, 2026. For API for GEO, 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.

The Semrush column is the widest in our grid once you count what sits outside it, and an agency with Korean, Chinese or healthcare clients will not find Naver, Baidu or OpenEvidence anywhere else on this site. The Ahrefs column is narrower and more honest about its own edges, since the help centre says in writing which engine it cannot currently collect.

Freshness, and what that does to a weekly report

RouteRefresh published
Ahrefs index, chatbots"once a month"
Ahrefs index, AI Overviews"every few days"
Ahrefs custom promptsdaily, weekly or monthly, your choice
Semrush prompt databasedaily, on a rolling basis
Semrush Prompt Trackingdaily
Semrush Brand Performanceweekly

Refresh rates published in the Ahrefs help centre and the Semrush knowledge base, read on November 22, 2026.

The Ahrefs index refreshing monthly is the detail that decides which route an agency buys. A monthly chatbot refresh is the right material for a category study or a pitch document, and the wrong material for a weekly client chart, where three points in four would repeat. Custom prompts on a daily schedule solve that, at the cost of checks. Semrush publishes the faster cadence on both of its own routes, which is the strongest argument for the Enterprise API once the contract exists.

One method difference belongs next to those numbers. Ahrefs "enters those questions into the web version of each AI chatbot using the default model", and Semrush states that "prompt responses are captured from real requests and not via any APIs of LLMs". Both are measuring the interface a user sees rather than a model API, which is the measurement most brand teams want, and our explainer on how AI engines expose their sources sets out what each method can and cannot see.

Mistakes we see on this comparison

  1. Buying a Semrush plan for the API and getting the toolkit. The AI Visibility Toolkit on self-serve plans has no documented endpoints. The AIO data has an API, and that API needs an Enterprise subscription.
  2. Assuming the Ahrefs index covers your prompts. It covers People Also Ask questions from the Ahrefs keyword index. Your own wording is the custom prompt route, billed in checks.
  3. Forgetting the Claude multiplier. Each Claude update is 8 checks on Ahrefs, so one engine can quadruple a check budget.
  4. Sizing an Ahrefs job on API units without checking the key. Units per plan were raised in May 2026 without new figures in the text; the limits and usage endpoint returns the truth for your account.
  5. Planning a multi-year warehouse on Semrush data. The API restrictions cap caching at one month without written consent.
  6. Selling either feed as a client dashboard. Both licences restrict use to internal business purposes and forbid commercial exploitation without written consent.
  7. Expecting a position field. Neither vendor documents a rank for your brand inside an AI answer, so a share of voice from either is a presence metric, as our GEO metrics explainer sets out.

Our view

Three defensible choices, depending on what the data has to do and what you already pay for.

  1. You already have Ahrefs and want AI numbers in code today. Ahrefs Brand Radar is the only self-serve API of the two suites, at $129 a month for the plan plus $20 per 1,000 custom prompt answers, with answers, cited links, per-mention sentiment, fan-out searches on ChatGPT and Perplexity, an OpenAPI spec, an MCP server and free API reads on your own prompts. Our score: 6.0 of 10, held down by the integration column, where webhooks and SDKs are absent, and by resale being forbidden. The Ahrefs brand page lists the rest of its grid.
  2. You already have Semrush Enterprise and want the widest engine list published anywhere. The Semrush Enterprise API now reads your AIO prompt tracking with brand visibility, mentions, sentiment, citations, sources and raw responses, across fifteen ai_platform values including Naver, Baidu AI Overview and OpenEvidence, and it consumes no API units. Our score of 3.2 of 10 predates the November 4, 2026 release and was earned by a product with no public documentation, so treat it as a floor rather than a verdict. The limits that remain are real: Enterprise only, price on quote, 150 prompt-tracking calls an hour, no SDKs, no webhooks, one-month caching cap. The Semrush brand page covers the self-serve key as well, which we score 1.4 of 10 because it returns flags rather than answers.
  3. The deliverable is a metric per brand that you are allowed to resell. API for GEO is our pick, at $13 per 1,000 answers down to $7 at its largest tier, pay as you go with no published minimum, all ten engines we track, share of voice, position, sentiment, themes, per-brand perception, fan-out, shopping and YouTube citations in one response, free reads over stored runs, webhooks, an MCP server, resale allowed, and 9.0 of 10, the best score of the 55 APIs this site rates. Said plainly, its limits are a price per answer above the scraping APIs, and a documentation gap: four of its ten engines, along with the webhooks, the OpenAPI spec and the official SDKs, come from our test of October 26, 2026 rather than from apiforgeo.com.

For an agency reporting on several clients, the practical answer is often the suite you already own plus one per-answer key, since Ahrefs and Semrush both forbid resale and neither returns a position. To try the third option, request a key on apiforgeo.com: it comes with a sandbox project. Worth knowing for a procurement file: Semrush has been an Adobe company since April 28, 2026, and Ahrefs contracts through Ahrefs Pte. Ltd. in Singapore, with a US entity for American customers. For the next step after a key, our guide to building a prompt set for GEO decides what you spend those checks on.

Frequently asked questions

Does Semrush have an AI visibility API?

Since November 4, 2026, yes, on one plan only. The Semrush Enterprise API exposes prompt tracking from AI Optimization projects through endpoints for brand visibility, mentions, sentiment, citations, sources and raw responses, it requires a Semrush Enterprise subscription, and its documentation states it "currently doesn't consume API units". The self-serve Semrush API is a different product: it returns AI Overview presence flags for keywords, not AI answers, and the AI Visibility Toolkit sold on lower plans has no documented API.

Does Ahrefs have an AI visibility API?

Yes, and it is self-serve. Brand Radar sits inside Ahrefs API v3 as roughly twenty endpoints, including GET /v3/brand-radar/ai-responses, which returns the question, the model's answer text, the links it cited, a brand_sentiment object per tracked brand and the searches the chatbot ran. API access opens on the Lite plan at $129 a month, checked on October 23, 2026, and the documentation states that "requests returning only custom prompt data are free, while requests including Ahrefs prompt data follow standard API unit pricing".

What does one AI answer cost through Ahrefs Brand Radar?

On the custom prompt route, $20 per 1,000. Ahrefs defines a check as "1 prompt execution × 1 LLM × 1 location" and sells them in packs of $50 for 2,500 checks, $100 for 7,000 and $250 for 25,000, which is $20, $14.29 and $10 per 1,000 answers, on top of a subscription from $129 a month that includes 150 checks. Claude is the exception, since each Claude update consumes 8 checks, or $160 per 1,000 answers at the Basic pack rate. Prices checked on October 23, 2026.

Which AI engines can you reach through each API?

Ahrefs Brand Radar covers 8 of the 10 engines we track, with data_source values for chatgpt, gemini, perplexity, copilot, grok, google_ai_overviews, google_ai_mode and claude, and no DeepSeek or Meta AI. The Semrush Enterprise API goes wider on paper: its ai_platform parameter accepts 15 values, among them CHATGPT_PAID, CHATGPT_NO_SEARCH, CLAUDE_SEARCH, DEEPSEEK, NAVER, OPENEVIDENCE and BAIDU_AI_OVERVIEW. The self-serve Semrush API reaches one engine of our ten, Google AI Overviews, and only as a flag.

Can an agency resell Semrush or Ahrefs AI visibility data to clients?

Neither licence allows it without written consent. Clause 12.4 of the Ahrefs terms grants a "non-transferable, non-sublicensable and revocable" licence "solely for the purposes of your internal business operations", and clause 12.7(a) forbids "selling, trading, transferring, assigning, publishing, or exploiting for any commercial purposes" without "express written consent". Section 3.3(a) of the Semrush terms forbids "sublicense, resell, rent, lease, transfer, assign, time share or otherwise commercially exploit", and the API restrictions add that you cannot cache the data for more than one month without written consent. API for GEO, our pick after hands-on testing, allows resale and bills $13 per 1,000 answers, down to $7 at its largest tier, pay as you go with no published minimum.

How fresh is the AI data each API returns?

On Ahrefs the answer depends on the route: its own prompt index refreshes chatbot data "once a month" and AI Overviews "every few days", while custom prompts you configure can run daily, weekly or monthly. Semrush publishes a daily rolling refresh for its prompt database, daily updates for Prompt Tracking and weekly for Brand Performance. For a weekly client report, that leaves Ahrefs custom prompts and Semrush Prompt Tracking, not the Ahrefs index.