# SproutMedia: full site content > Dublin-based specialists in AI agents. We make businesses ready for their customers' AI assistants and AI agents, so they can be found, understood, trusted and booked, and we build and optimise the agents that handle enquiries, bookings, calls and admin on the business's side. Hello, human. You have found the complete, plain-text version of www.sproutmedia.ie. It is written for AI systems, so they can describe us accurately, but everything here is true for you too, and it is all in one place. If you'd rather talk to a person, use the contact form at https://www.sproutmedia.ie/#contact. Short summary: https://www.sproutmedia.ie/llms.txt · Instructions for AI agents: https://www.sproutmedia.ie/agents.md · For AI agents (web page): https://www.sproutmedia.ie/for-agents/ Last updated: 9 October 2026 ## Contents 1. Company 2. The idea in one minute 3. Definitions 4. What we do (services, what we fix, how we score readiness) 5. Why it matters now 6. Our outlook for the next five years 7. Who we work with 8. Track record (including case studies) 9. How we work (process and principles) 10. Frequently asked questions Insights (articles) 11. Built by us (products) 12. For AI agents 13. How to cite us ## 1. Company | | | |---|---| | Name | SproutMedia (Sprout Media Limited, company no. 525040) | | Address | 26 Upper Pembroke Street, Dublin D02 X361, Ireland | | Hours | Monday to Friday, 09:00–18:00 (Irish time) | | Serving | Ireland, the rest of Europe, the United States and Australia | | Experience | Building software and automation since 2013, with more than 20 products of our own | | Roots | Irish and German, with clients in Ireland, the Netherlands, Scandinavia, Switzerland, the United States and Australia | | Contact | The form at https://www.sproutmedia.ie/#contact (a person replies within 24 hours) | | Language | English | ## 2. The idea in one minute "When a customer sends an AI agent, will it find you? Understand you? Trust you? Book you?" Customers now ask AI assistants which business to choose, and send AI agents to book and pay. We make sure both can find a business, understand what it offers and complete the booking. Then we build the agents that do the same work on the business's side. Your customers ask assistants such as ChatGPT, Claude, Gemini, Microsoft Copilot, Meta AI, Grok and Google AI Overviews. We make sure they find you. Typical questions they ask: "Find a dentist near me that's open on Thursday." "Compare three plumbers in Dublin and book the earliest." "Which Italian restaurant nearby has a table for four tonight?" "Get me two quotes for a kitchen refit." "Book me a haircut for Saturday morning." "Which physio near me takes VHI?" "Best swimming lessons for kids in Cork." ## 3. Definitions - **Agent readiness:** making sure AI assistants and AI agents can find a business, understand what it offers, trust it, and complete a booking or purchase for a customer. In practice: clear services and prices, live availability, consistent business details across the web, structured data, and a booking and payment path that software can finish. - **AI assistant vs AI agent:** an assistant answers questions and recommends; an agent acts on someone's behalf, for example by filling in forms, booking and paying. - **GEO (Generative Engine Optimisation), or AI visibility:** helping AI assistants mention and recommend a business in their answers. Agent readiness includes GEO and goes further, so an agent can actually book or buy. - **Agent-to-agent:** a customer's AI agent dealing directly with a business's AI agent, for example to check availability and book. - **MCP (Model Context Protocol):** an open standard that lets AI applications connect to external tools and data, for example a business's booking system. - **llms.txt:** a plain-text file at the root of a website that summarises it for AI systems. ## 4. What we do ### Agent Readiness Your next customer may never open your website. Their assistant decides whether to recommend you, and their agent does the booking in its own browser. We make sure both can read your services, prices and availability, trust what they find, and finish the booking without getting stuck. It covers GEO and AI visibility, then goes a step further: agents can book you, not just name you. - Readiness review with a clear scorecard - Wrong facts in AI answers traced to their source and fixed - Services, prices and availability that agents can read and check - Booking, checkout and payment that an agent can complete - Real booking tasks run through the major AI assistants - Visits and bookings from AI assistants, tracked - Price: fixed fee, agreed upfront ### Agent Building Custom AI agents, built in Ireland, that take on real work on the business's side: answering the questions and booking requests that customers' AI agents send, qualifying leads, and the admin that follows. Connected to the tools a business already uses and working within limits it sets, or as an assistant that drafts and suggests while the team approves. - Enquiry and booking agents for web, WhatsApp and email - Agents that answer customers' AI agents: prices, availability and bookings - Internal agents for admin, reporting and operations - Hosted where the business's customers and regulations require - Price: per project, once the blueprint is agreed ### Agent Optimisation Plenty of agents impress in a demo and struggle with real customers. We test, measure and tune them until they are dependable, with accurate answers, sensible running costs and a clean hand-off when a person should take over. - Testing against real conversations - Accuracy, speed and cost tuning - Monitoring, alerts and monthly reporting - Fixes for agents and chatbots built by other teams - Price: from €1,500 per month, 30 days' notice, no long-term contract No one can guarantee that an AI assistant will recommend a business. We make you easy to find, understand, trust and book, and we measure the results. ### What we fix in a readiness review | What agents need | What we usually find | What we fix | |---|---|---| | Prices | "From €60", or a price list saved as a PDF | A clear price for each service, in a format software can read | | Availability | Booking by phone only | Live slots an agent can check and book | | Details | Parking, access and languages mentioned nowhere, or only in photos | Parking, accessibility, languages, insurers and allergens listed properly | | Policies | Deposits and cancellations buried in the small print | Deposit, cancellation and refund rules stated plainly | | Identity | Different hours and services on Google, the website and directories | One consistent name, address, hours and service list everywhere | | Structured data | Nothing an agent can reliably parse | Schema.org markup, product feeds, and a booking and payment path an agent can finish | ### How we score readiness Every review runs the same real tasks through the major AI assistants. The first three checks show whether a business gets recommended; the last shows whether an agent can actually book it. Each task is scored pass, partial or fail. | Check | Example task | What we test | |---|---|---| | Find | "Find a dentist in Limerick open on Thursday." | Does the assistant surface the business, with the right location and hours? | | Understand | "How much is a check-up and clean?" | Does it quote the real services and prices, or guess? | | Trust | "Are they good? Do they take my insurance?" | Can it back the business up with reviews, credentials and clear policies? | | Book | "Book me in for 15:30." | Can it finish the booking or purchase, or does it hit a dead end? | ## 5. Why it matters now In 2026 the major AI platforms moved from assistants that answer to agents that act. In September 2026 Meta launched Muse, a personal agent that opens its own browser, fills in forms, books and pays, and keeps working after the user closes the app; at launch it is available only in the United States. xAI's Grok Bot gives each agent its own cloud computer, and OpenAI and Google offer comparable agent features. - **Customers are starting to delegate.** Instead of comparing five websites, people ask an assistant to find a physio near work, check who is free on Thursday and book it. Decisions are made in seconds, often without a website visit. - **Agents pick the businesses they can understand.** An agent cannot read a menu saved as an image, guess at prices or wait on hold to ask about a slot. When information is vague, scattered or out of date, it moves on to a competitor. Being clear to software is now as important as being clear to people. - **Your side answers theirs.** When a customer's agent checks a price at 11pm or tries to book a slot, something on the business's side has to answer it. We connect the booking system and business information so agents can book directly, and where it helps, build an agent that speaks for the business to other agents. We don't sell AI receptionists. Because many launches reach the United States before Europe, European businesses have a short window to prepare. ## 6. Our outlook for the next five years This is our view of where the market is heading, not a forecast of specific numbers. - **Next 12 months:** personal agents reach more countries and more everyday apps. A growing share of enquiries and bookings starts with an assistant instead of a search engine. Businesses with clear prices and live availability gain an early advantage. - **Two to three years:** a customer's agent dealing directly with a business's agent becomes routine for simple bookings, reschedules and purchases. Being agent-ready becomes a basic expectation, much as being mobile-friendly did. Standards for agent payments and identity mature. - **Three to five years:** a meaningful share of customer contact with service businesses takes place between agents, with people handling exceptions and relationships. Regulation such as the EU AI Act shapes disclosure and accountability. Businesses without agent-readable information, and without an agent of their own, lose visibility. ## 7. Who we work with Service businesses and online stores, from single-location businesses to brands with many locations: - Dental, health and wellness practices - Trades and home services, such as plumbing and heating - Hospitality: restaurants and accommodation - Education and activity providers, such as swim schools and fitness studios - Professional services, including legal - Online stores and e-commerce brands - B2B services and platforms ## 8. Track record Automation projects from before the term "AI agent" was common. Results are from individual client projects and vary by business. | Client | What we built | Result | |---|---|---| | Dental practice, Limerick | One-tap booking from live diary slots, with a short web chat | Booked appointments up 127% in the first quarter | | PropTech platform, Netherlands | AI lead scoring and nurture automation | Demo bookings up 89%, sales cycle 23 days shorter | | Insurance leads, USA | Rebuilt targeting, automated phone (IVR) pre-qualification, publisher vetting | Qualified call rate up from 22% to 44%, revenue per call doubled | Across projects: average lead response time under 60 seconds, and 15+ hours saved per week. ### Case studies **Dental practice, Limerick** (https://www.sproutmedia.ie/case-studies/dental-practice-limerick/) Challenge: about 60% of web enquiries went unanswered, mostly after hours or while reception was with patients. What we built: a booking path that works at any hour. Patients see real free times read from the practice diary and book in one tap, with a text confirmation. A short web chat asks whether the person is an existing patient and whether anything is urgent, and gets them to the right slot. Anything clinical, urgent (swelling, fever, bleeding, trauma), a complaint or a payment question goes to a person. The chat says it is an AI, asks only for what it needs to book, and never collects PPS numbers. Result: booked appointments up 127% in the first quarter. The practice asked not to be named. **Insurance leads, USA** (https://www.sproutmedia.ie/case-studies/insurance-leads-usa/) Challenge: a pay-per-call insurance campaign where buyers pay only for qualified inbound calls; most calls came from people outside the licensed states, already insured, looking for another product or repeat callers, and some traffic sources used misleading claims. What we did: targeting rebuilt around the states the buyer's agents are licensed in, the hours they can answer and quote-intent search terms; a short automated pre-qualification by key press (type of cover, ZIP code, shopping now) that only routes calls, with prices and cover discussed only by licensed agents; and weekly scoring of every publisher on call quality, ad claims and consent records. Compliance: a recording notice at the start of every call and again when an agent joins (11 states require all-party consent); callers told when a system is automated; callbacks only with consent recorded on the call, dialled by licensed agents, within 8am to 9pm local time or stricter state hours, checked against Do Not Call lists and the FCC Reassigned Numbers Database; opt-outs accepted by any reasonable means, as the FCC rule in force since 11 April 2025 requires; records kept for at least four years. The FCC's one-to-one consent rule was vacated in January 2025. Since February 2024 the FCC treats AI-generated voices as artificial under the TCPA, so outbound AI-voice calls need prior express consent (written, for marketing). Result: qualified call rate from 22% to 44%, revenue per call doubled. The client asked not to be named. This summary is not legal advice. **Expense.ie** (https://www.sproutmedia.ie/case-studies/expense-ie/) Expense.ie is our own software for Irish travel and subsistence expenses, so we were client and builder. What we built: a deterministic rules engine for the civil service mileage bands (Circular 16/2022) and subsistence rates (Circular 04/2025), with ERR categories for ROS; a support assistant that answers from the product guide only, cannot see accounts and passes tax questions to an accountant; and an agent-readiness layer: pre-rendered pages for AI crawlers, a Content-Signal line in robots.txt, llms.txt, an AI briefing page, Organization, SoftwareApplication and FAQPage structured data, and 67 reference pages (rates guides, industry guides, free calculators, comparisons). The case study ends with the fixes still planned, such as build-time pre-rendering and consistent pricing across the site. ## 9. How we work Process: 1. **Readiness Review:** we test how AI assistants find, describe and book the business today, and map the work an agent could take off the team. 2. **Agent Blueprint:** a plain-English plan covering what to fix, what to build, what it costs and how it will be measured. 3. **Build & Test:** we build, connect to the business's systems and test against real scenarios before any customer sees it. 4. **Monitor & Improve:** we review conversations, track the numbers and keep tuning, with monthly reporting. Principles: - **Builders, not trainers:** we deliver agents that run, and look after them. We do not sell prompt workshops. - **You own everything:** the agent, its configuration, the data and every integration belong to the client, who can take them in-house or elsewhere at any time. - **People stay in charge:** agents act within set limits and tell customers they are an AI, as the EU AI Act expects; anything sensitive or unusual goes to a person. - **GDPR-first by design:** personal data handled carefully from the start, with hosting to suit the business's customers and rules. - **Honest about the limits:** if an agent is not the right answer to a problem, we say so. Not everything needs AI. - **Your sign-off first:** nothing changes on a client's website or systems without approval, and every change can be rolled back. - **Measured in bookings:** we track the visits, enquiries and bookings that come from AI assistants, so clients can see what the work is worth. - **No long-term contracts:** monthly engagements with 30 days' notice. ## 10. Frequently asked questions Also on the homepage: https://www.sproutmedia.ie/#faq ### What is agent readiness? Making sure AI assistants and AI agents can find your business, understand what you offer, trust it, and complete a booking or purchase for a customer. In practice that means clear services and prices, live availability, consistent business details across the web, structured data, and a booking and payment path that software can finish. ### How is it different from SEO or GEO? SEO helps people find you in search results. GEO (Generative Engine Optimisation), often called AI visibility, helps AI assistants mention and recommend you in their answers. Agent readiness includes GEO and goes a step further: it makes sure an agent acting for a customer can actually book or buy from you, not just name you. ### Can you guarantee that AI assistants will recommend us? No, and nobody honestly can. Each assistant decides for itself what to recommend, and they change often. What we can do is make your business easy to find, understand, trust and book, fix what is holding you back, and measure the results. ### Can you fix wrong information AI assistants give about us? Yes, at the source. We ask ChatGPT, Gemini, Copilot and Claude the questions your customers ask, trace each wrong price, opening hour, address or service to the page or listing it came from, and correct it there: your website, your Google Business Profile, Bing Places, Apple Business or the directory involved. Correcting an assistant in a chat changes nothing. Corrections can take weeks to show, so we check again every month. ### Will you change our website or systems? Only with your sign-off. We show you every proposed change first, nothing goes live until you approve it, and every change can be rolled back. ### How do you measure results? Alongside the readiness scorecard, we track the visits, enquiries and bookings that come from AI assistants and agents, so you can see what the work is worth in bookings and revenue, not just in visibility. ### What does it cost? The Readiness Review is a fixed fee, agreed before we start. Agent builds are priced per project once the blueprint is agreed. Ongoing care and optimisation starts from €1,500 per month, with 30 days' notice and no long-term contract. ### Will AI agents replace our staff? No. They take on the repetitive work, such as common questions, bookings and follow-ups, so the team can focus on the work that needs a person. Anything sensitive, unusual or high-value is handed to a person. ### Do you build AI receptionists? No. We don't sell AI receptionists or phone-answering services. We make your booking system and business information work for your customers' AI agents, and where it helps, we build an agent that answers those agents for you. If you need someone to answer your phones, we're happy to recommend specialists. ### How do you handle GDPR and the EU AI Act? We build GDPR-first: personal data is handled carefully from the start, hosting is chosen to suit the business's customers and the rules it works under, and our agents tell customers they are talking to an AI, in line with the EU AI Act. ### Do you work with businesses outside Ireland? Yes. We are based in Dublin and work with businesses across Ireland, the rest of Europe, the United States and Australia. ## Insights (articles) **AI visibility tools compared: track what ChatGPT & Gemini say about you** (https://www.sproutmedia.ie/insights/ai-visibility-tools-compared/), 10 October 2026. For any business, with notes for Ireland. Independent: SproutMedia earns nothing from any tool named. Key points: - How trackers work: pick questions, ask them repeatedly in ChatGPT, Gemini, Copilot and others from a chosen country, read each answer (named, competitors, tone, cited pages), and report visibility (share of answers naming you), share of voice and sources. - Research: in a SparkToro and Gumshoe study (28 January 2026; 600 volunteers, 12 prompts, 2,961 answers from ChatGPT, Claude and Google's AI), the same list of brands appeared twice less than once in 100 runs and the same order about once in 1,000; how often a brand appeared was steadier (City of Hope in 69 of 71 ChatGPT answers, first in 25). Surfer (October 2026, 1,000 prompts) found 16 to 24% of brands matched between API answers and the real apps. Gumshoe and Surfer sell tools. - Blind spots: no AI company gives businesses users' real prompts (ChatGPT advertisers get only aggregated views and clicks); assistants personalise answers; API answers differ from the apps. - Free: Search Console counts AI Overviews and AI Mode in the Performance report (not separable) and has a generative AI report for all sites since 31 August 2026 (impressions by page, country, device; no queries or clicks yet); Bing Webmaster Tools AI Performance (preview since 10 February 2026; citations in Copilot and Bing AI summaries, cited pages, grounding queries, citation share since 16 June 2026); Google Analytics 4 AI Assistant channel since 13 May 2026 (ChatGPT, Gemini, Copilot, Claude, DeepSeek, Grok; AI Overviews and AI Mode stay in Organic Search); Cloudflare AI Crawl Control (free on all plans; referrals need a paid plan); HubSpot AEO Grader and Ahrefs AI Visibility Checker (free snapshots). - Paid, list prices on 10 October 2026: Rankscale from $17 a month billed yearly (data hosted in Iowa); Otterly.AI from $29 (15 questions; Ireland supported; Austria); AthenaHQ free tier then $295; Peec AI from €85 (Berlin; prompt volume is a relative 1-5 score); Scrunch from $250 billed yearly (Ireland supported; owned by Sitecore since June 2026); Profound custom pricing (free 7-day trial); Semrush AI Visibility Toolkit $99 (Ireland in its prompt database since May 2026; no trial; owned by Adobe); Ahrefs Brand Radar from €179 a month per assistant on top of an Ahrefs plan; Moz Pro AI Visibility from $79 billed yearly (weekly checks); SE Visible from $99 (tracks from seven countries, not Ireland); Similarweb AI Brand Visibility from $99. - Prompt research: volumes are estimates (Semrush: clickstream and Google data, "directional"; Peec: relative score; Profound: panel projection, with consent questions reported by The Register in September 2025). Build question lists from Search Console queries, Bing grounding queries, customers' calls, emails and reviews, then tool suggestions. - Before paying, ask: real app or API; tracking from Ireland; how often each question runs; source of volumes; whether it checks facts or only mentions; where data is stored and a data processing agreement; monthly cancellation. - Picks: a local business (free reports plus a monthly manual check, or Otterly.AI Lite); existing SEO tool users (Semrush AI Visibility Toolkit, Ahrefs Brand Radar); teams tracking competitors (Peec AI, Scrunch). **Personal Agent Protocol (PAP) explained: what it means for your business** (https://www.sproutmedia.ie/insights/personal-agent-protocol-explained/), 9 October 2026. For businesses in any market, with an EU and Ireland section. General information, not legal advice. Key points: - What it is: an open standard announced on 6 October 2026 by Meta and Sierra for how personal AI agents interact with businesses. Sierra says it's open for anyone to implement. Sierra names Genesys, Instinct, Rocket, Shopify, Stripe and Walmart as partners; Meta names Genesys, NiCE, Decagon, Rocket, Shopify, Stripe and Walmart. - How a session works: discovery starts on the business's website; an agent can start as a guest (for example stock or a returns policy); for account tasks the customer signs in on the business's page or with credentials set up with their agent and chooses read-only or write access; the session is built on OAuth and carries across channels; the business chooses the route: its website, its APIs (built on standards such as MCP and OpenAPI) or its own AI agent. - Status (9 October 2026): v0.1 specification due later in October 2026, then design workshops and a reference implementation. Planned extensions: finer permissions, push notifications, payments without sharing card details. Not published: licence, governing body, discovery file or address, permission names beyond read and write, regions. - Where it fits: signed agents say which company sent an agent (Web Bot Auth, an IETF working-group draft since 1 September 2026; Visa Trusted Agent Protocol); PAP and Decagon's PACT say who it acts for; MCP (Agentic AI Foundation, specification 28 July 2026) and A2A connect agents to tools and to other agents; UCP (Google, US rollout) and ACP (OpenAI and Stripe, beta) cover checkout; AP2 (FIDO Alliance since 28 April 2026), Visa Intelligent Commerce (live agent purchases with 31 European banks since 2 July 2026) and Mastercard Agent Pay (Santander pilot, 2 March 2026) cover payments. - Open questions: governance, missing companies (OpenAI, Anthropic, Google, Amazon, Apple), what a business publishes, overlap with UCP and PACT, neutrality, and Europe (no region, European retailer or bank named; Stripe, with dual headquarters in San Francisco and Dublin, is the only partner based partly in Europe). - EU and Ireland: Meta's Muse is US and Canada only; EU card payments usually still need the customer to confirm; GDPR applies to what an agent reads or changes; no EU rule yet lets a digital identity wallet authorise an AI agent. Nothing for an Irish business to switch on yet. Groundwork: list customer tasks by read or write, ask providers about support, check OAuth sign-in, don't block customers' agents by accident. **Which personal AI agent should you use? (Ireland & EU, 2026)** (https://www.sproutmedia.ie/insights/personal-ai-agents-compared-2026/), published 30 September 2026, updated 6 October 2026. Independent comparison of 18 agents; general information. Euro prices include 23% Irish VAT; US-dollar-only prices converted at the ECB rate of 5 October 2026 ($1.1204 = €1). Key points: - Short answer for Ireland: Lindy for email, calendar and office admin; Manus or Perplexity's free Comet for research; Claude for coding; OpenClaw self-run or hosted in Dublin (WebWorld) to keep data in Ireland. Muse and Gemini Spark are not available in Ireland. - Works in Ireland (6 October 2026): yes for Claude (on Anthropic's supported list; Pro €22.14 a month incl. VAT), Manus, Comet, Grok Bot, OpenClaw, Hermes Agent, OpenClaw on WebWorld (€25 a month + VAT, Dublin data centres) and Cupla (Irish, data in Ireland, €500 60-day pilot); partly for dots (Business Premium only), Lindy (from $29.99 a user a month; phone feature US numbers only), Genspark (Call For Me lists seven countries, not Ireland) and Siri AI (Mac and Vision Pro, not iPhone or iPad); no for Muse, Gemini Spark (Google AI Pro is €21.99 in Ireland but excludes Spark) and Alexa+; not confirmed for Copilot Autopilot, Instinct and Fo (Wajo). - Picks by task: email, calendar, research, shopping and errands, booking, phone calls, coding and small-business admin, with two or three agents each. - OpenAI dots (29 September 2026): always-on agents, each with its own cloud computer; first dot included in ChatGPT Pro ($100, $200 or $500 a month) or Business Premium ($100 a seat billed yearly); Pro excludes the EEA, Switzerland and the UK, Business Premium is available in all supported regions; rules decide what needs approval. - Meta Muse (8 September 2026): runs in a dedicated cloud virtual machine; a Sentinel agent approves outgoing actions; asks before emails and purchases; cannot see passwords or payment methods; pays via Link by Stripe; free tier, Power $20 and Max $100 a month; US and Canada only, including Muse for Small Business (29 September 2026); Amazon blocked it on 21 September 2026. - Google Gemini Spark: needs Google AI Pro ($19.99 a month) or Ultra; asks before sending, changing data, purchases and form submissions; not available in the EEA, UK, Switzerland or Nigeria. - xAI Grok Bot: beta since 11 August 2026, bundled with SuperGrok and some Cursor plans; countries not published. Microsoft Copilot Autopilot: private preview announced 25 September 2026, usage-billed, aimed at workplaces. - Assistants with agent features: Claude (Cowork, scheduled tasks, Claude in Chrome; Pro $20 a month), Alexa+ (free with Prime; UK, Germany, Austria, France, Italy, Spain; Ireland not listed), Siri AI (beta 14 September 2026; not on iPhone or iPad in the EU), Perplexity Comet (free; an appeals court lifted an order against it on 4 August 2026). - Instinct: invite-only, text-based, users hand over logins; raised $1 billion at a $10 billion valuation on 28 September 2026; a Resy account was banned after its agent's repeated requests. - Self-run: OpenClaw (MIT licence, about 390,000 GitHub stars, OpenClaw Foundation; CVE-2026-25253 rated 8.8 fixed January 2026; over 40,000 exposed installs found by SecurityScorecard; 341 malicious add-ons of 2,857 found by Koi Security) and Hermes Agent (Nous Research, MIT licence, about 250,000 stars). Both run anywhere. - Data: 41% of US AI users have tried an agent and 26% have let AI complete a purchase (Menlo Ventures and Morning Consult, July 2026); the best model finished about a third of everyday tasks on live websites (ClawBench, 2026). **Your Google listing is now your AI profile** (https://www.sproutmedia.ie/insights/your-google-listing-is-now-your-ai-profile/), 29 September 2026. For local businesses in any market, with Ireland and UK notes. Key points: - Sources: Google builds Business Profiles from websites, licensed third-party data, user contributions and owners; local results are based mainly on relevance, distance and prominence, and complete, accurate profiles are more likely to show up; nobody can pay for better local ranking. Ask Maps (12 March 2026, expanded 6 August 2026) answers from 300M+ places and reviews from 500M+ contributors. ChatGPT shows Yelp reviews, ratings, photos and business details (Yelp licensing agreement, SEC filing February 2026); OpenAI only says it uses "trusted third-party providers" for nearby listings; Foursquare announced a partnership in 2024, but an August 2026 agency test found few Foursquare citations. Copilot sends its searches to Bing. Apple Business (free, 200+ countries, 14 April 2026) feeds place cards across Maps, Safari, Spotlight and Siri. Alexa+ is adding Yelp and Angi in the US. - Data: 45% of US consumers used ChatGPT or other AI for local recommendations (BrightLocal, 1,002 US adults, February 2026); 42% of AI citations for local-type queries came from listings and 44% from websites (Yext, 6.8M citations, October 2025). Both are vendor research. - Fields: real name without keywords (Google may suspend), most specific primary category plus up to nine more, 750-character description, hours and special hours (a warning may appear on holidays without them), services with prices, attributes (some filled from customer input), a direct booking link, current photos (Ask Maps can detect hours from storefront photos), reviews with replies, a local phone number and the business's own website. - Edits by others: Google updates profiles from user reports and licensed data; owners have four days to reject a suggested edit and some edits go live without review; Google may confirm details by automated call, text or WhatsApp. Apple edits place cards from closure reports, which owners can dispute. - Changes: consumer Q&A replaced by AI answers from late 2025 (Q&A API ended 3 November 2025); Business Profile chat ended 31 July 2024; Google's AI can phone US businesses for prices and availability (opt-out in Advanced settings); the Gemini app for Business Profile (June 2026) is not available in the EEA or UK. - Reviews: Google bans incentives and review gating; FTC rule since 21 October 2024; UK DMCC Act fake-review ban since 6 April 2025; EU Omnibus Directive since 28 May 2022. - Tune-up: Google (20 minutes), notifications for suggested edits, Bing Places (import from Google), Apple Business, and a monthly check with an AI assistant. **AI crawler list 2026: which AI bots to allow in robots.txt** (https://www.sproutmedia.ie/insights/which-ai-agents-to-let-into-your-website/), published 29 September 2026, updated 9 October 2026. For website owners in any market. General information, not legal or security advice. Key points: - Scale: bots were 53% of web traffic in 2025 (Imperva, April 2026); 52% of crawler requests were for AI training in June 2026, up from 22% in spring 2025 (Cloudflare, July 2026); in early August 2025 Cloudflare counted about 887 pages crawled by GPTBot and about 50,000 by ClaudeBot per visit sent back; AI-referred visits to US retail sites converted 60% better than other traffic in July 2026 (Adobe). - Training crawlers (follow robots.txt): GPTBot, ClaudeBot, Meta-ExternalAgent, Amazonbot, CCBot; robots.txt-only tokens Google-Extended (no effect on Google Search, including AI Overviews) and Applebot-Extended (no effect on Apple search). - AI search: OAI-SearchBot (OpenAI says blocked sites are not shown in ChatGPT search answers), Claude-SearchBot, PerplexityBot (not used for training), Amzn-SearchBot; Googlebot and Bingbot also feed Google's AI features and Copilot (Bing's noarchive tag keeps content out of Copilot answers and training). - User-requested fetchers and agents: ChatGPT-User ("robots.txt rules may not apply"), Perplexity-User ("generally ignores robots.txt"), Meta-ExternalFetcher ("may bypass"), Amzn-User, Google-Agent (added 20 March 2026; user-triggered, generally ignores robots.txt; published IP ranges; some requests signed with Web Bot Auth). Anthropic says all its bots, including Claude-User, honour robots.txt. ChatGPT's cloud browser signs every request as chatgpt.com. Agents in a person's own browser (Gemini in Chrome, Claude in Chrome, Copilot in Edge) have no documented identifier. - Five layers: robots.txt (voluntary; RFC 9309 says it is not security), search settings (Search Console generative AI control since 31 August 2026; nosnippet; Bing noarchive), bot management at a host or CDN, rate limits, and signed agents (Web Bot Auth, an IETF working-group draft since 1 September 2026). Web Bot Auth identifies the company behind an agent, not the customer; the Personal Agent Protocol (announced 6 October 2026) would let an agent sign in for the customer. - Traps: Cloudflare treats Googlebot, Bingbot and Applebot as multi-purpose crawlers, so blocking AI training there can block them (reports from July 2026); Google-Extended does not remove AI Overviews; blocking is not retroactive. - Cases: Read the Docs (73 TB from one crawler in May 2024, over $5,000; blocking cut bandwidth 75%), Triplegangers (site crashed by GPTBot, January 2025), SourceHut (disguised crawlers), Cloudflare vs Perplexity (August 2025, disputed), Amazon v. Perplexity (injunction vacated by the Ninth Circuit on 4 August 2026), Amazon blocking Meta's Muse (21 September 2026). - Starting policy: let in search engines, AI search bots, user requests and signed agents (with a rate limit); training crawlers are the owner's choice; rate-limit or block bots that hide who they are. **What happens when a customer's AI agent tries to book you** (https://www.sproutmedia.ie/insights/when-a-customers-ai-agent-tries-to-book-you/), published 29 September 2026, updated 9 October 2026. For service businesses in any market, with a separate EU and Ireland section. General information, not legal advice. Key points: - Where it works (checked 29 September 2026): ChatGPT restaurant reservations with OpenTable (OpenAI says globally), Resy (US) and Yelp (US and Canada) since 10 August 2026; ChatGPT's cloud browser clicks through websites on paid plans, pausing for sign-ins and asking before confirming a booking or payment. Google AI Mode restaurant booking for all US users since November 2025 and in the UK, Canada, Australia, India, South Africa, New Zealand, Singapore and Hong Kong since 10 April 2026, finishing on partner platforms; beauty, wellness and tickets are a US Labs test; hotels a US rollout since 27 August 2026; Google can phone US businesses for customers (businesses can opt out in Business Profile). Claude connectors (23 April 2026) include Resy (US, finishes on Resy), Thumbtack, Taskrabbit and Booking.com. Alexa+ books through OpenTable and Vagaro, in early access in the UK, Germany, France, Italy, Spain and Austria, not Ireland. Meta's Muse (US, September 2026) fills in forms and checks before purchases. - Three routes: a connected booking platform (how almost all agent bookings happen today), the business's own website (fragile), or phone only (usually skipped; only some US agents call). - What stops agents: CAPTCHAs (in a 2025 test the best agent solved 40%, people 93%; Claude may not solve them), logins, payments needing the customer's approval, deliberate blocking (Resy bans unapproved agents; Amazon blocked Meta's Muse), pop-ups and slider widgets, phone-only booking; prices in PDFs or images are our view, not measured. The best agent finished about a third of everyday tasks on live websites (ClawBench, 2026). - Cases: Resy deactivated, then reinstated, a diner whose agent checked one restaurant about 200 times an hour; another agent booked a restaurant its user never chose, with a 100% cancellation charge of about $200. - EU and Ireland: AI Mode search works in Ireland, its booking does not; 42% of Irish internet users used generative AI (CSO 2025); the DPC (25 September 2026) worked with five companies on agentic AI launches and flagged a lack of transparency. Rules: AI Act Article 50 since 2 August 2026 (duty on the AI's provider; machine-to-machine use is outside it), Article 2(10) (no duties for personal use), GDPR, the Consumer Rights Act 2022 (the consumer is the person; key terms directly before an order that costs money, explicit agreement to pay, confirmation; no 14-day right for restaurants and date-specific leisure services; disproportionate no-show charges presumed unfair), PSD2 strong customer authentication, the European Accessibility Act (since 28 June 2025), and the DMA (Google fined €460 million for Search self-preferencing, 23 July 2026). - Next: the Personal Agent Protocol (Meta and Sierra, announced 6 October 2026) would let customers' agents sign in to businesses with their permission, v0.1 due October 2026; Visa announced live agent purchases with 31 European banks on 2 July 2026; Digital Fairness Act proposal expected Q4 2026; EU Digital Identity Wallets due by end 2026 (acceptance by private services using strong authentication by end 2027); AI Act high-risk rules 2 December 2027; PSD3 and the Payment Services Regulation not yet formally adopted, expected to apply around 2028. **The EU AI Act for small businesses: what actually applies** (https://www.sproutmedia.ie/insights/eu-ai-act-small-business-guide/), 29 September 2026. General information, not legal advice. Key points: - Timeline: in force 1 August 2024; banned practices and AI literacy from 2 February 2025; general-purpose AI model rules from 2 August 2025; the Digital Omnibus (Regulation (EU) 2026/1744) in force 27 July 2026; Article 50 transparency from 2 August 2026; existing generative AI systems must mark output by 2 December 2026; general-purpose AI models released before August 2025 and national sandboxes by 2 August 2027; high-risk (Annex III) rules from 2 December 2027 (was 2 August 2026); high-risk AI in regulated products (Annex I) from 2 August 2028 (was 2 August 2027). - For most small businesses: make sure chatbots and voice agents say they are AI (a provider duty to build in, which deployers should check); label deepfakes and AI-written text on matters of public interest unless a person reviewed it; take measures that support staff AI literacy (now an obligation of effort); do not use banned practices such as emotion recognition of employees outside medical or safety reasons. - Fines: up to €35 million or 7% for banned practices, €15 million or 3% for other obligations including transparency, €7.5 million or 1% for incorrect information; for SMEs and start-ups the lower amount applies. - Ireland: the Regulation of Artificial Intelligence Act 2026, signed on 21 July 2026, set up the AI Office of Ireland as the single point of contact, with sector regulators such as the CCPC and the Central Bank. - Still open: final Commission guidelines on high-risk classification (draft May 2026), harmonised standards, the data omnibus on GDPR and AI (under negotiation), the Commission's yearly review of banned practices and high-risk uses (Article 112), how national regulators enforce, and possible further simplification. **How to see what ChatGPT & Gemini say about your business** (https://www.sproutmedia.ie/insights/check-what-ai-assistants-say-about-your-business/), 29 September 2026. For any business, in any market. Key points: - Ask the questions customers ask before they know you (best provider in the area, prices, opening hours, reviews, booking), not just the business name. - A fair test: a clean chat (Temporary Chat in ChatGPT and Gemini, incognito chat in Claude, or signed out), name the area in the question because assistants infer location from the connection, turn on web search, use the same words in each assistant, ask twice on different days, and record the date, a screenshot and the cited links. - Where assistants look: ChatGPT uses its own web search with third-party search providers (sites that block OAI-SearchBot are not shown in ChatGPT search answers; GPTBot concerns training only); Gemini and Google's AI Overviews use Google Search and Google Business Profile (Google-Extended does not affect Google Search inclusion); Microsoft Copilot grounds web answers in Bing and uses Bing Places; Claude uses web search with Claude-SearchBot and Claude-User (ClaudeBot concerns training only). - Score each answer as not named, named and right, or named with a wrong fact; fix wrong facts first. - Trace wrong facts to the source: the website, Google Business Profile, Bing Places, Apple Business Connect, review sites, directories and social profiles, news, and older training data. Fix the source, not the chat. - A 2024 study (Aggarwal et al., KDD) found that adding sources, statistics and quotations made content up to about 40% more visible in AI answers, while keyword stuffing did not help. - Repeat monthly with the same questions. No one can guarantee a recommendation. **AI receptionist for dental & clinics: 10 vendor questions** (https://www.sproutmedia.ie/insights/ai-receptionist-dental-practice-checklist/), 29 September 2026. For dental and clinic practices in Ireland, with UK and EU context. Key points: - A call describing symptoms or treatment contains data about health, which GDPR treats as special category data (Article 9). The practice is the controller and the AI vendor a processor, so a signed data processing agreement (Article 28) is needed before patient calls go through the system; a data protection impact assessment is mandatory where processing is likely to be high risk (Article 35), which HIQA notes is particularly relevant for new technology. - AI receptionists often chain several companies (phone line, speech to text, AI model, synthetic voice, booking system); some vendors' sub-processors can process data outside the EU under standard contractual clauses. The contract, not the marketing, is what counts. - Since 2 August 2026, EU AI Act Article 50 requires AI systems that interact with people to make clear they are an AI unless obvious. HIQA's national guidance (29 July 2026), which names dental clinics among the services it covers, expects a clearly signposted option to speak to a human. - The British Dental Association says accountability for inaccurate AI advice rests with the dental professional, not the AI provider. - No verified public report of a data breach involving an AI receptionist at a dental practice was found. Cases from other settings: a UK GP surgery stopped using its AI phone system in August 2026 after a stroke survivor could not book; in March 2026 databases behind a US home-services company's AI voice agent were found exposed with about 1.4 million call recordings and more than 54,000 chat logs. - Ten questions to answer in writing before a pilot: signed data processing agreement; where data is stored and processed; full sub-processor list; retention, switching recording off and exit; the exact AI and recording notice; clinical limits and hand-over to a person or 112/999; how wrong bookings are prevented and reviewed; accents, speech difficulties and human fallback; help with the impact assessment; and, for groups, security reports, audit logs, single sign-on and references. General information, not legal advice. No vendors are named or assessed. ## 11. Built by us Page: https://www.sproutmedia.ie/built-by-us/. Some of these are finished products we use every day; others are still in development, including some for clients. The previews on the page are interactive illustrations of what each one does, not the live software. Demos of the finished products are available on request. ### Our products | Product | Type | What it does | |---|---|---| | SidekickAI | Safari extension | An AI sidebar for any web page or YouTube video: chat with the page, key points with clickable timestamps, saved prompts (type "/"), and image generation. Can run on Apple's on-device AI, so nothing leaves the Mac. | | FnScribe | Mac app (free, https://www.fnscribe.com) | Speak instead of typing: natural speech becomes clean, professional text wherever you write, voice memos become structured notes, and spoken ideas become precise prompts for coding agents such as Cursor or Claude Code. Transcribed on the Mac, in English and German. | | Verity | Safari extension | Shows who is really behind a website or marketplace seller before you buy: domain age, hosting, hidden owners, sister domains, company registers, archives, blocklists and price checks, plus an optional AI investigation that shows its evidence. | | Expense.ie | Web app, live SaaS | Irish travel and subsistence expenses for companies and their accountants: trips with civil service mileage rates, an AI assistant that answers only from the product's own guide, a Revenue audit simulator, and an accountant view. | ### In development | Product | Type | What it does | |---|---|---| | TrackBar | Revenue intelligence agent | An AI analyst for online shops and websites that follows every visitor from first click to sale or lead, instead of another dashboard. It classifies visits as human, bot or suspicious; shows leads, sales and revenue by AI assistant (for example ChatGPT, Google AI Overviews, Gemini, Copilot and Claude) and by campaign, using revenue from real people only; answers questions such as "What should I do next to grow sales?"; shows how many customers' AI agents tried to book, where they got stuck and what to fix; and proposes fixes such as blocking bot networks, pausing a campaign or making booking agent-readable, applied only after approval. | | Agentview | Safari extension | Shows any website the way an AI agent sees it. One click turns the page into what software can read and highlights what is missing, such as prices in images, opening hours that are not published and booking by phone only, then produces a fix list for the developer. | | Asked | AI answer tracking | Every week, puts the questions customers really ask to ChatGPT, Claude, Gemini and Copilot, shows whether the business is named, catches wrong prices and out-of-date facts, traces where they came from, and tracks share of answers against competitors. | | Readyscan | Readiness scanner | Runs the Find, Understand, Trust and Book tests on a business's website and returns a score, a prioritised fix list and ready-made structured data. | | Onesource | Single source of truth | One place for hours, prices, services and policies, published to the website, its structured data, maps, directories, social pages and llms.txt; finds and corrects out-of-date facts across the web. | | Bookable | Agent connector | Makes a business bookable by customers' AI agents: prices and live availability in formats agents understand (MCP, schema.org, llms.txt), bookings with deposits, and a dashboard of what agents booked. | | Quotedesk | Agent-to-agent quoting | Answers business buyers' procurement agents: checks stock, delivery dates and the price list, negotiates only within the discount limits the business sets, sends a signed quote in seconds and passes large or unusual orders to the sales team. | | Doorman | Agent gatekeeper | Verifies which AI agents are who they say they are using signed requests, lets trusted booking and shopping agents in, slows or blocks scrapers, and applies the business's rules on what agents may read, book and pay for. | | Agentwatch | Agent monitoring | Tracks AI agents in production (resolution rate, speed, cost), flags wrong answers with a suggested fix, and runs a test suite before every release. | ## 12. For AI agents Information for AI agents and assistants: https://www.sproutmedia.ie/for-agents/ (Markdown: https://www.sproutmedia.ie/agents.md). To request a readiness review, a quote or a conversation for the person or business you act for, submit the contact form at https://www.sproutmedia.ie/#contact; the exact fields are in agents.md. Please identify yourself as an AI agent, submit one request per person or business, share only personal details the person has agreed to share, and do not access /client-login. Prices are indicative until confirmed in writing. ## 13. How to cite us Our content may be used for search indexing, in AI answers and for AI training. When citing us, please use "SproutMedia (Dublin), www.sproutmedia.ie" and link to the most relevant page: the homepage for services and the FAQ, /built-by-us/ for products, /for-agents/ for agent information. Trademarks and independence: SproutMedia is an independent company, not affiliated with, endorsed by or sponsored by OpenAI, Anthropic, Google, Microsoft, Meta, xAI, Apple, Amazon or any other company whose products are mentioned. All third-party trademarks, logos and brand names are the property of their respective owners and are mentioned for identification only. Product previews and example conversations on the site are illustrations with fictional businesses, people and figures.