There is now an entire consulting category selling you a fix for this, most of it invented in the last eighteen months. Some of what they sell is fine. Some of it is provably useless. Here’s how to tell the difference.
Pew Research Center, surveying US adults in February 2026, found that about half — 49% — now use AI chatbots, with 44% using ChatGPT specifically. That’s roughly a doubling of ChatGPT use since 2023.
BrightLocal’s 2026 consumer survey found 45% of consumers said they had used AI to get a local business recommendation. One survey, one methodology — treat it as a signal, not a measurement.
And the traffic side, which is the part nobody selling AEO wants to discuss: Pew’s 2025 browsing-behavior study — 900 US adults, 68,879 Google searches in March 2025 — found that visits where an AI summary appeared ended in a click on a traditional result 8% of the time, versus 15% without a summary, and that links inside AI summaries were clicked on just 1% of visits.
Read that honestly. AI answers are not a new traffic channel. They are a channel where being mentioned is the win, and the click may never come. If a vendor’s pitch is “AI traffic,” ask them which number they’re forecasting.
Google published explicit guidance on optimizing for its generative AI features. The headlines:
That includes llms.txt, the file half the AEO industry has been billing for. No major AI company has documented using it as a ranking or retrieval input. I keep one on my own site because it’s a tidy plain-English summary of the business and costs nothing, but I’d be lying if I sold it as an AI visibility strategy, and so is anyone who does.
The practitioner research doesn’t agree with itself either. Ahrefs and Semrush have published conflicting findings on how much AI citations overlap with traditional search rankings. When the two largest tool vendors in the field can’t agree on the basic correlation, nobody has a formula.
1. Be crawlable by the right bots. If you want to be cited, you have to be readable. These are the user agents that matter, and they do different jobs:
GPTBot (training), OAI-SearchBot (search index), ChatGPT-User (live fetch when a user asks)Google-Extended (Gemini / AI training; separate from Googlebot)ClaudeBot, Claude-User, Claude-SearchBotPerplexityBot, Perplexity-UserThe distinction is worth a minute of thought. You can decline to feed model training while still allowing the search and live-fetch bots that get you recommended. Blocking everything is the most common own goal on this list.
2. Write the page that answers the question. Models cite sources that contain a clear, self-contained answer near the question. A page titled with the actual question, answering it in the first two sentences, then supporting it — that’s the format. This is also just good writing, which is why Google keeps insisting it’s still SEO.
3. Put your facts where they can be read. Hours, service area, what you do and don’t do, pricing structure, who you serve. In text, on the page, not baked into an image or trapped in a PDF. An AI cannot recommend what it cannot parse.
4. Get mentioned somewhere other than your own website. The consistent pattern across every credible analysis is that models lean on third-party sources — directories, local press, forums, review sites — to decide who’s real. Your own claim that you’re the best carries no weight anywhere, with anyone.
5. Keep your Google Business Profile accurate. Local answers still lean heavily on it. That’s covered in the previous post.
Notice what isn’t on this list: any file, any markup, any keyword density trick. The work is having verifiable facts in readable places and a reason for someone else to mention you.
One adjacent thing, since it sits in the same bucket for most owners.
Estimates of how many small businesses actually use AI vary wildly depending on who’s asking and how they define it — Census Bureau data from May 2026 found fewer than 20% of firms with four or fewer employees reported using AI, while industry surveys from the Chamber of Commerce and Intuit have reported 60% and 68%. That spread isn’t a mystery; it’s a definitional gap. “Used ChatGPT once” and “runs AI in production” are not the same claim.
Which matters because the FTC’s Operation AI Comply, announced in September 2024, is an enforcement sweep specifically targeting deceptive AI claims — companies overstating what their AI does. The same standard applies to a five-person business on a landing page as to a startup with a Series B.
Say what you actually run. It’s a shorter sentence, and it’s the one you can defend.
So this post is built from two sources only: Google’s own documentation, and the Whitespark 2026 Local Search Ranking Factors report, which surveys local SEO practitioners and is the closest thing the field has to a shared evidence base. Where they disagree, I’ll say so.
Google’s own guidance on local ranking is short enough to read in five minutes, and it says this:
“Local results are mainly based on relevance, distance, and popularity.”
Relevance is how well your profile matches the search. Distance is how far you are from the searcher. Prominence is how well-known the business is — and Google is explicit that “more reviews and positive ratings can help your business’s local ranking.”
Google also says, plainly: “There’s no way to request or pay for a better local ranking on Google.” Anyone selling you guaranteed map-pack placement is selling you something Google says does not exist.
And: “Businesses with complete and accurate info are more likely to show up in local search results.” Which is the least exciting and most reliable advice in local SEO — fill it in, all of it, correctly.
Whitespark’s 2026 report ranks the factors its respondents believe drive local pack and Maps results. Their top five for the local pack:
For local organic results — the blue links under the map — the top factors look completely different, and this is the finding most small businesses miss:
Three practical readings of that list.
Your primary category is the single highest-leverage field in the entire profile. One dropdown. If you’re a plumber whose primary category is “Contractor,” you are competing in the wrong race. Pick the most specific category that describes your main service, and use secondary categories for the rest.
Hours are a ranking factor, not just a courtesy. “Business is open at the time of search” sitting in the top five means stale holiday hours cost you visibility, not just goodwill.
One page per service. A single “Services” page listing eleven things loses to eleven pages. This is the highest-value website work most local businesses never do, and it’s the thing you should do instead of obsessing over the profile once the profile is complete.
This part matters more than any optimization, because a suspended profile is worth zero regardless of your category choice.
1. Your business name must be your real-world business name. Not “Plumber Dallas — 24/7 Emergency Plumbing.” Google’s naming guidelines require the name you actually use in the real world, on your signage and paperwork. Keyword stuffing the title is an effective tactic right up until it isn’t, and the penalty is your listing.
2. Never gate reviews. Google’s prohibited content policy is explicit that businesses must not “discourage or prohibit negative reviews, or selectively solicit positive reviews from customers.” Routing happy customers to Google and unhappy ones to a private form is exactly that. Ask everyone, the same way, every time — and don’t pressure people on the premises either, which the same policy also covers.
3. Photos should be your actual business. Google’s photo and video policy requires imagery that genuinely represents the business and complies with its content policies. I’d add a practical rule on top: don’t post a generated image of a storefront, van or team you don’t have. Even where a policy line is arguable, a customer who arrives to a place that looks nothing like the photos has already decided something about you.
LocalBusiness schema markup on your website is not a documented map-pack ranking lever. Google’s structured data documentation describes what the markup does — it tells Google about hours, departments and the like, and can change how your business appears in results — and Google explicitly does not guarantee that features consuming structured data will show at all. Meanwhile Google’s own local ranking guidance never mentions markup: it lists relevance, distance and prominence. Google has also said structured data isn’t required for its AI features. Add it because it’s tidy and cheap. Don’t add it expecting to move up the map pack.
If you do nothing else this month:
That list will beat almost any agency retainer sold on the promise of a ranking Google says can’t be bought.
Most owners treat this as a character flaw in customers. It usually isn’t. People book things three weeks out, on a phone, between other tasks, and then life happens to them. The businesses with low no-show rates aren’t serving better people. They built a system.
Appointment reminders are one of the few operational tactics with proper randomized-trial evidence behind them, because healthcare has been studying the problem for decades.
A 2013 Cochrane systematic review of mobile phone messaging reminders for healthcare appointments found attendance improved from roughly 67.8% with no reminder to about 78.6% with a text reminder. That is an enormous effect for something that costs a fraction of a cent to send.
A separate 2013 randomized trial published in BMC Health Services Research compared text reminders against telephone reminders in a primary care clinic and found text messaging performed comparably to phone calls — at substantially lower cost.
Two honest caveats. This is healthcare, where the appointment is usually months out and often unpleasant; your trade appointment may not behave identically. And a reminder can’t fix a booking someone regretted the moment they made it. But the direction is well-established and the mechanism is obvious: most no-shows are forgetting, not deciding.
Not “more reminders.” A specific, small number, each doing a different job.
Immediately on booking — the confirmation. This is the most-read message you will ever send, and most businesses waste it. It should contain: date and time in plain words (“Thursday, October 2, at 10:00am”), where, who is coming, roughly how long, what they need to have ready, what it costs or what the deposit covers, and how to reschedule in one tap.
24 hours before — the reminder. Short. Time, place, and a reschedule link. That’s it.
A couple of hours before, for a visit-based service — the “on my way” window. “Marcus will be there between 10 and 11.” This one kills the anxious cancellation, where somebody bails because they don’t know if you’re actually coming.
Three messages, each with a job. A fourth is nagging, and nagging trains people to ignore you.
Here is the shift that changes the number: make rescheduling easier than not showing up.
Right now, for most businesses, a customer who realizes at 8am that they can’t make 10am has two options: call and have an awkward conversation with a person, or just not be there. The second option costs them nothing socially. So they take it, and you learn about it by standing in a driveway.
Put a one-tap reschedule link in every message. You are not “making it easy to cancel.” You are converting a silent no-show into a rebooked job and an open slot you can still fill.
Shorten the gap. No-show risk grows with lead time. If you’re booking three weeks out, that’s a capacity problem wearing a no-show costume — and a waitlist that can fill a hole on two hours’ notice is worth more than a fourth reminder.
Take something. A deposit, a card on file, or even a required confirmation tap changes behavior, because it converts a vague intention into a commitment. Match the size to the job: a $250 deposit on a $9,000 install is normal; a deposit on a $60 consult reads as distrust and costs you bookings. Pick one, say it clearly at booking time, and never surprise anyone with it.
Reply to the reply. If a reminder sends from a number nobody monitors, people who try to reschedule get nothing back and become no-shows anyway. A reminder system that can’t receive a message is half a system.
Two numbers:
You want the first low and the second healthy. A business with a 0% reschedule rate and a 14% no-show rate has made rescheduling too hard. Moving those customers from column one to column two is the cheapest revenue available to you this month.
I want to be careful here, because this topic is a swamp of invented statistics. You will find confident claims that a specific percentage of small business calls go unanswered, and a specific dollar figure lost per year. I could not trace those to a real study, so I’m not going to repeat them, and you should be suspicious of any agency that does.
Here’s what’s actually documented.
Invoca’s Call Conversion Industry Benchmarks Report, 2025 analyzed over 60 million phone calls across nine industries. Two findings matter for a small business:
Sit with the second one. A call that reaches a human converts at a rate that no other channel in a small business touches. Roughly a third of the time, the sale closes inside the conversation. Which makes the first number the expensive one.
And it’s Invoca’s own data from their own customers — businesses sophisticated enough to be paying for call analytics. It’s fair to assume the long tail of small businesses does worse, not better.
If you relied on Google to catch calls for you, note that Google discontinued the call history and chat features in Google Business Profile as of July 31, 2024 — in its own words, “As of July 31, 2024, the chat and call history features are no longer available in your Business Profile.” Calls placed from your Google listing now just go to your phone, and if nobody picks up, there’s no record inside Google to go back to. Anything you had there is gone.
That’s worth saying out loud because a lot of owners still believe Google is quietly holding their messages.
1. A voicemail greeting that’s a promise, not an apology. Most greetings say the business is unavailable and to leave a message. Better: say when you call back and give the alternative. “You’ve reached [business]. We’re on a job right now. Leave your name and number and you’ll get a call back before end of day — or text this same number and we’ll usually answer faster.” That’s a free change that converts a dead end into two live paths.
2. Let people text the number they’re already calling. Many people would rather text. If your business line can receive texts, say so on the website, in the greeting, and on the Google listing.
3. Route after hours somewhere honest. Forwarding to a personal cell that gets answered while driving is worse than voicemail. Pick one.
4. Make a rule for callbacks. Not “when I get to it.” A named window. The previous post on this blog covers why the window matters more than the polish.
5. Then, and only then, consider AI answering.
I build AI systems for a living, so read this with that in mind: an AI phone answerer is a real tool and it is oversold.
What the evidence says about customer appetite: Gartner’s 2024 consumer survey found that 64% of customers would prefer that companies didn’t use AI for customer service, and 53% said they would consider switching to a competitor if they found out a company was going to use AI for customer service. That’s not an argument against AI. It’s an argument against hiding it, and against putting it between a ready-to-buy caller and a human.
There are also hard legal edges, and they’re moving fast:
So the version I’d actually build for a small business:
That system is worth building. A system that tries to be you on the phone isn’t — customers notice, the Gartner numbers say what they think about it, and the legal exposure lands on you, not the vendor.
For two weeks, log every inbound call: time, whether a human answered, and whether it turned into work. You don’t need software. A notes app is enough.
Most owners are genuinely surprised by the result, in both directions. Some find the phone is fine and the real leak is somewhere else. Others find an entire afternoon of the week where nobody is home. Either way, you stop guessing — and you stop paying consultants who quote statistics they can’t source.
That is the most expensive moment on your website, because you paid for everything that came before it. And it’s the moment most owners never look at, because nothing shows up in the inbox to look at.
The best-documented version of this is ecommerce checkout. The Baymard Institute maintains a rolling meta-analysis of documented cart abandonment studies; the current average sits around 70% of carts abandoned. Their large-scale checkout usability research points at causes that have nothing to do with price: being forced to create an account, and a checkout that is simply too long or too complicated, each drive a meaningful share of abandonment. Baymard’s benchmarking has found the average checkout runs to roughly 5 steps and 11 form fields, where their usability testing suggests around 8 fields is achievable for most.
A booking form is a checkout. Same psychology, smaller stakes, usually worse design — because a checkout gets optimized by a team and a booking form gets built once by whoever set up the website.
Two more numbers worth carrying around:
Go to your own booking form on your phone. Count the fields. Then, for each one, ask a single question: can I do my job without this answer right now?
A useful benchmark: name, one way to reach them, and one sentence about what they need. Everything else is a question you can ask a person who has already said yes.
Here is where most “cut your fields” advice gets dishonest.
In a well-documented test written up by Michael Aagaard at CXL, removing form fields reduced conversions. The shortened form pulled in more submissions of lower quality and less intent; the longer form’s extra questions acted as a qualifier and a signal that the business took the work seriously.
So the rule is not “fewer fields always wins.” The rule is: every field must earn its place by either helping the customer get served or helping you sort real work from noise. A question that qualifies is worth keeping. A question that only feeds a spreadsheet nobody reads is not. If you sell a $12,000 job, a couple of serious questions protect your calendar. If you sell a $90 appointment, they don’t.
Test it if you can. If you can’t run a real test — and most small businesses can’t get enough traffic for a clean one — default to short, and add a question back only when you can name the decision it changes.
If your form works and inquiries still evaporate, the leak has moved downstream — to how fast anyone replies, or to whether the appointment gets kept. Those are separate problems with separate fixes, and the booking form usually gets blamed for both.
Fix the form first anyway. It’s the one part of the customer path you fully control, and it’s an afternoon’s work.
This is the least glamorous idea in marketing, which is exactly why it keeps working.
The number everyone quotes comes from a 2007 study by James Oldroyd, then at Kellogg, run with the lead-response software company InsideSales. Looking at inbound web leads, it found that the odds of making meaningful contact with a lead dropped sharply as the wait grew — contacting a lead within five minutes versus thirty minutes produced dramatically different odds of ever reaching that person. It’s a vendor-funded study from almost two decades ago, so treat the exact multiplier as folklore. Treat the direction as real, because it keeps replicating.
In 2011, Harvard Business Review published an audit of 2,241 US companies’ actual response behavior. The findings were blunt: only 37% responded to a lead within an hour, and 23% of the companies never responded at all. That’s not a conversion problem. That’s a business quietly refusing money.
A 2014 InsideSales study of a different set of companies landed in the same place — 47% of the companies studied never responded to the inquiry. Different researchers, different years, same embarrassing result.
Here is the part owners underrate: your competitor’s response time is your real benchmark, not perfection. If four businesses get the same inquiry and three of them take a day, you don’t have to be instant. You have to be first.
Nobody decides to ignore a lead. The gap opens for ordinary reasons:
Notice that none of these are fixed by working harder. They’re all structural. Structure is fixable.
1. Separate new inquiries from everything else. A new lead should never land in a general inbox. Route it to its own destination — a dedicated address, a channel, a text to your phone. If a new customer’s message looks like every other notification, it gets treated like every other notification.
2. Send an honest auto-reply in under a minute. Not “Thank you for contacting us, we value your business.” That reads as an unmanned building. Something closer to: “Got it — this came through at 6:42pm. I’ll reply personally before 9am tomorrow. If it’s urgent, call or text this number.” Specific, human, and it makes a promise you can keep. An auto-reply is not a response; it’s a placeholder that buys you the night.
3. Decide the owner in advance. One named person owns new inquiries during each block of the day. Not a team. A person.
4. Set a floor, not a fantasy. “Every inquiry gets a real human reply within one business hour” beats “we aim to respond quickly” because you can tell whether you did it.
5. Keep following up more than once. The other half of the HBR and InsideSales findings is that most businesses that do respond, respond once. One email, no answer, done. Three touches over five days — email, a call, a short text — is not pestering. It’s the normal cost of doing business with a human who has a job and children.
You need a dashboard of exactly two things:
Pull them once a week. If the median is in hours, you have room. If it’s in days, you have a hole in the boat that no ad budget will patch.
Speed doesn’t rescue a bad fit. If someone wants a service you don’t sell, answering in four seconds just gets you to “no” faster — which is still worth something, because it frees the slot. And there are businesses where a same-day reply is genuinely enough; a wedding photographer booked eleven months out is not in a five-minute race. The principle isn’t be instant. It’s close the gap between the moment someone decides to act and the moment a human confirms they were heard. For most small businesses, that gap is the single largest unforced loss in the whole customer path.
It’s also the cheapest thing on this website to fix.
Here are four that hold up when things are loud.
When a child is overwhelmed, the instinct is to solve the problem fast. But naming the feeling first — “you’re really frustrated that the tower fell” — does something quiet and powerful: it tells the child that the feeling is allowed, and that you see it. Research on emotional labeling suggests that putting words to a feeling can take some of the heat out of it. You’re not agreeing the meltdown is reasonable; you’re just lowering the temperature enough to think.
“Take a deep breath” is abstract to a four-year-old. Give the breath a shape instead. Trace a finger up one side of the other hand on the inhale, down on the exhale — a slow five-finger trace. Or pretend to smell a flower, then blow out a candle. The body leads the mind here: a longer exhale gently nudges the nervous system toward calm, and a child who can see the breath is far more likely to actually take it.
Big feelings come with a flooded brain. Anchoring attention to a single concrete thing gives it somewhere to land. “Find three things in this room that are blue.” “What’s the quietest sound you can hear right now?” It feels almost too simple, but narrowing focus is exactly what a flooded mind needs — one small thing instead of everything at once.
The hardest time to learn a new skill is mid-meltdown. So practice the pause when nobody needs it — as a game before bed, a silly breathing contest, a thirty-second “reset” you do together after dinner. Then, in the hard moment, you’re not teaching something new; you’re reaching for something familiar.
The goal was never a perfectly calm child. It’s a child who knows, somewhere in their body, that the storm passes — and that someone will sit with them until it does.
That’s the whole idea behind Pausely: not another thing demanding attention, but a small, shared moment of calm you can reach for together. If that resonates, say hello — we’d love to hear how your family handles the big-feeling moments.
]]>