With GBP Q&A gone and AI answers pulling from reviews and websites, what has actually changed in how you optimise profiles for service businesses?

1 reply

• 

  adminAdmin

  Original poster 18 hours ago

  The short answer is that local pack ranking has changed much less than the answer layer on top of it. What you lost with Q&A was the ability to write the answer to a customer’s pre-sale question yourself, directly on the profile. Google now writes that answer with Gemini, using your profile fields, your reviews, your photos and your website. So the work has moved from “fill in every profile field” to “make sure the evidence behind each likely answer exists, is current, and agrees across every source Google reads”.

  For a service business the shift is bigger than for a restaurant or a shop. A service-area business usually hides its address, has few attributes and no menu, so there are fewer structured facts for the AI to use. That pushes more of the load onto review text and your service pages. How much this matters to you depends on how many questions customers ask before they call. It matters a lot for a termite inspection or a solar install. It matters less for an emergency locksmith, where proximity, open hours and a phone number still do most of the work.

  What actually changed in 2025 and 2026

  The timeline matters because the rollout was uneven. Whitespark’s Miriam Ellis documented that Google deprecated the Q&A API in late 2025 and the Q&A section then disappeared from profiles. Businesses can no longer add new questions and answers. Google launched Ask Maps in March 2026 in the US and India. It then expanded in August 2026, in English, to Australia, Canada and more than 150 other countries and territories. Several of the newer agentic features in Ask Maps launched in the US first, so check what is live in your market before building a plan around it.

  Area	Before	Now	What it means for a service business
  Pre-sale questions	Owners could seed and answer Q&A on the profile	Gemini answers questions from profile data, reviews, photos and web content	The answer has to exist somewhere Google reads. You cannot pin it to the profile any more.
  Reviews	Mostly read as rating, count and recency	Also summarised by AI and cited inside answers	What customers write about the job matters, not just the stars
  Owner replies	Written manually	Google is testing AI-suggested replies (limited test, March 2026)	Generic replies are easier to produce and still read as generic
  Profile edits by users	“Suggest an edit” form	Conversational edits inside Ask Maps (US first)	Check your hours and details more often
  Map pack ranking	Category, proximity, name, reviews, hours	Broadly the same, based on expert opinion	Do not drop the fundamentals to chase AI answers

  On that last row: Whitespark’s 2026 Local Search Ranking Factors survey (47 experts, published November 2025) still ranks primary category, proximity, keywords in the business name, physical address in the city and open hours as the top pack factors. It is an expert opinion survey, not a controlled study. Keywords in the name also carry a trap: Google’s profile guidelines say the name should reflect your real-world name, and that adding unnecessary information “could result in the suspension of your Business Profile”. That risk has not changed.

  How an AI answer about your business gets built

  Q&A worked like a noticeboard outside your shop. You could pin up the answers you wanted people to see, and most people never checked them. The new system works more like a reporter who has been asked a customer’s question. The reporter reads your listing, skims your reviews, looks at your photos and checks your website, then writes a two-line answer. You cannot hand the reporter a script. You can make sure the facts are easy to find and do not contradict each other.

  Technically, this is retrieval followed by synthesis. Google says Ask Maps draws on more than 300 million places and reviews from over 500 million contributors, but it has not published how it weights each source. Ellis’s test of the “ask a question” feature returned an answer that cited reviews and the business’s uploaded menu. Her article lists profile details, photos, videos, website content and some third-party sources as inputs. Treat that as a well-informed observation, not a documented ranking of sources.

  A customer asks: “Do they do same-day hot water repairs in Northside?”

  ↓

  Google retrieves candidate evidence: services list, hours, service area, reviews that mention the job, photos, website pages

  ↓

  Do the sources clearly and consistently answer the question?

  ↓

  Yes: a specific answer, often citing a review. No or conflicting: a vague answer, a hedge, or an answer built from an old or wrong source.

  The practical shift:

  The unit of optimisation is now the customer question, not the profile field. A complete profile that never mentions the job the customer asked about still produces a weak answer.

  What I would change on a service business profile now

  1. Turn the old Q&A into a question bank, then answer it where Google reads

  Collect the questions people actually ask before booking. Call recordings, enquiry forms, quote emails and whatever you had in Q&A are better sources than keyword tools. Typical service questions cover price ranges, callout fees, lead times, areas covered, licences and insurance, warranties, and whether you handle a specific job type. Answer them on the relevant service pages in plain language. Mirror the short version in your profile’s services list and description where the fields allow it. Google’s guide to generative AI features says you do not need special markup, llms.txt files or AI-specific rewording for this. Clear, crawlable pages are the requirement.

  2. Treat review text as content, and get it without breaking policy

  Reviews now do two jobs: persuading the person reading them and supplying facts to the AI. A review that says “great service” answers nothing. A review that says “replaced our hot water system the same afternoon in Northside” answers three questions. BrightLocal’s 2026 consumer survey (1,002 US adults) found 82% of consumers read AI-generated review summaries and 74% look for reviews from the last three months. That supports asking every customer promptly and consistently.

  Watch for this:

  Google’s Maps content policy prohibits offering incentives for reviews and says businesses must not “selectively solicit positive reviews from customers”. Ask everyone. Use an open prompt such as “What did we do for you, and how did it go?” instead of handing customers keywords to include. Scripted, look-alike reviews are a policy risk and give the AI less real detail to work with.

  3. Fix the themes the summaries will surface

  AI review summaries aggregate recurring complaints. If six reviews mention late arrivals, the summary may say so, and no profile edit will write over it. That is an operations problem wearing SEO clothing. Replies still matter for human readers. The same BrightLocal survey found half of consumers are unlikely to choose a business that uses generic templated responses. If Google’s AI-suggested replies reach your account, edit them so they address the specific job and complaint.

  4. Show the work in photos and video

  Before-and-after shots, the crew on site, branded vehicles and close-ups of the finished job give Google visual evidence of what you do. Stock images add nothing. For service-area businesses, photos taken across different suburbs also show where you actually work.

  5. Keep the facts tight and check them more often

  Hours, service areas, categories and services need to match your website. Ask Maps now lets users suggest edits conversationally (US first), so a wrong holiday-hours change can now come from a customer’s chat as well as the old edit form. A monthly check is sensible for most businesses.

  6. Prepare for Google phoning you

  At I/O 2026, Google said Search can call businesses on a user’s behalf in select categories, naming home repair, beauty and pet care. The rollout was set for US users in summer 2026. If you are in those categories in the US, answering the phone and giving a clear price range or availability becomes a visibility issue, not just a sales one. There is no public evidence yet on how Google chooses which businesses to call.

  7. Remember that not every AI answer comes from Google

  OpenAI’s help documentation says ChatGPT search sometimes sends queries to partner search providers, and it names Microsoft. It also says ChatGPT may share a user’s location with third-party providers to show nearby business listings. Your Google profile is not the only record that matters. Consistent listings on Bing Places, Apple Business Connect and the main directories in your industry are cheap insurance. Experts in the Whitespark survey rated directory listings and expert-curated “best of” lists highly for AI visibility. That is opinion, but it matches how retrieval systems work.

  Hypothetical example:

  A pest control company covers three regions. A customer asks Ask Maps whether it treats termites in a particular suburb and offers a warranty. The profile lists only “Pest control service”. The website has one general services page, and most reviews mention cockroaches and ants. The likely result is a vague answer, or a competitor with a termite page gets the recommendation. The fix is a termite service page with warranty terms and the areas covered, a termite service in the profile, photos from real inspections, and review requests sent to recent termite customers without telling them what to write. Then ask the same question again and record the answer.

  How to tell whether any of this is working

  Measurement is the weakest part of this. As far as I can verify, Google has not documented a Business Profile metric that isolates Ask Maps exposure. On the website side, Search Console’s generative AI performance reports show impressions only, with no clicks or queries. They began with a subset of sites and reached all eligible sites worldwide on 31 August 2026. GA4’s AI Assistant channel catches some ChatGPT, Gemini and Copilot referrals, but traffic without a referrer still lands in Direct. You will need to observe the answers directly.

  A small controlled test is realistic:

  1. Hypothesis: adding explicit answers to the website and profile raises the share of priority questions that Ask Maps answers correctly and specifically.
  2. Setup: pick 20 real customer questions. Split them into a test group of 10 and a control group of 10 with similar difficulty.
  3. Baseline: ask each question three times, on a clean device, from inside the service area. Record the answer, whether it is correct, and which sources it cites.
  4. Change: fix the evidence for the test group only.
  5. Recheck: at four and eight weeks, using the same method.
  6. Metrics: the primary metric is the percentage of questions answered correctly. Secondary metrics are profile calls, website clicks, enquiry volume, and whether callers already know you do the job.

  Answers vary with personalisation, location and model updates, so expect noise. If the test group improves and the control does not, roll the approach out. If both groups move, the model probably changed, not your inputs. If neither moves, look at which sources the answers actually cite. They may be third-party pages you have not touched.

  Where I would put the effort

  If this were my profile, I would keep the ranking fundamentals exactly as they were and spend the time Q&A used to take on three things: a question bank answered on the website, a steady flow of specific and unscripted reviews, and fixing the complaints the summaries keep repeating. Then I would ask the AI my customers’ real questions every month and treat each wrong or vague answer as a missing piece of evidence to add.

  Need help with this on your own site?

  Niraj Raut works with businesses in Nepal, Australia, the UK, Europe and the US on technical SEO, ecommerce SEO, local SEO and AI search.

  See SEO services