Your map pack ranking hasn't moved all year, and your traffic is still down. Those two facts stopped contradicting each other sometime in 2025.
Google now answers many local research questions on the results page itself, naming businesses as it does so. Those aren't always the ones at the top of the map pack, because the overview gets assembled from a different set of sources.
I run link building and outreach at Hunter, so I spend most of my week looking at which pages get cited and why. Before writing this, I ran ten local research queries across eight verticals and logged every source in every citation strip. What came back contradicted most of the advice I'd read, and some I'd repeated myself.
Transactional searches were mostly fine. Research searches were not.
"Plumber near me" still lands people in the map pack. What got hollowed out is the layer above it: "do I need to replace or repair my furnace," "how much does Invisalign cost in Toronto," "what to look for when hiring a personal injury lawyer." Those queries brought someone to your blog post three weeks before they called you.
Two numbers frame it. Google searches ended without a click 68.01% of the time in the US during the first four months of 2026, up from 60.45% in 2024. And 45% of consumers now use AI tools to find local businesses, against 6% a year earlier.
One honest caveat: "transactional is safe" is getting repeated more confidently than the data supports. Semrush’s tracking shows AI Overviews on transactional queries climbing from 1.98% to 13.94% over ten months. Safer, then. Not safe.
So you didn't lose local rankings. You lost the top of your funnel. The person who read three articles before phoning now reads one paragraph.
I checked 60 citations. Most of them were competitor websites.
The standard advice is that AI Overviews run on third-party validation, so you should chase citations, directory listings, and "best of" roundups. I believed that until I checked.
Ten queries, eight verticals, all Toronto-area. Sixty citations in total.
Source types across 60 AI Overview citations, September 2026. Source: Graph from author's own audit.
The headline is the first row. Three quarters of what Google quoted was somebody's own commercial website, most often a dental practice, a law firm, a roofer or a physio clinic.
The absence is just as loud. Across sixty citations, there wasn't one "best of [service] in [city]" listicle or city guide. One review platform appeared, HomeStars, on a single query. If you've been buying directory placements to win AI Overviews, you bought the wrong thing.
"How much does Invisalign cost in Toronto" returned six citations, and all six were dental practices competing for the same patients.
Every source Google quoted on this query is a competing clinic. Source: Screenshot by user on Google Search,
20 September 2026.
Two patterns explain which sites got in. Every cited page answered the exact question rather than a broader version of it, and almost all put a specific number in plain text instead of behind a quote form. The Invisalign pages cited were the ones carrying a cost table.
There's a limit here, and I'd rather state it than have you find it. Ten queries in one metro is a sample, not a study, and AI Overviews are generated per location, so your own audit may not match mine. Run your own; it takes an hour, and the answer is specific to you.
Write the page that answers the question, with the number in it
If competitor sites are taking three-quarters of the citations, the work is making yours the one worth quoting.
Start with the basics, because none of the rest survives without them. Google Business Profile complete, LocalBusiness and Service schema, and location pages that answer a real question each instead of running one template with the city name swapped. If you want the full version of that groundwork, the tactics for the evolving Google Business Profile guide covers it properly.
My audit points at something simpler than all of that. Publish the number.
Every Invisalign page Google quoted had a cost table. Every furnace page had a price range by efficiency tier. The pages that got cited weren’t the most authoritative sites in the vertical; they were the ones that answered "how much" with an amount instead of "contact us for a quote." Google is also now asking AI to check prices for local businesses, which makes a number trapped inside a JPEG worth nothing.
The rule is narrow and testable. Take your ten highest-intent research questions, write one page for each, and put the answer in plain text near the top. A range is fine; a range with the variables explained is better, and vagueness is what keeps you out.
That's ten pages of genuinely specific writing, more than most local businesses manage in a quarter. If you're outsourcing it, brief on the question and the number rather than a word count, and check that whoever writes it will do the pricing research. Marketplaces like Legiit are one route, a local freelancer who knows the trade is another, and the brief matters more than either.
Reviews are a body of text, not a number
The star rating is a filter. The review text is what gets quoted.
BrightLocal’s 2026 survey puts a figure on it. Asked what makes a review matter, consumers ranked "backed up by other reviews with similar sentiment" first at 56%, while "the review has a high star rating" came fourth at 42%. Agreement across many reviews beats the number attached to them.
Which makes your review request script a content brief, and most of them are wasted. "Please leave us a review" produces "great service, highly recommend," and fifty of those describe nothing anyone can quote. Ask for the specific service and the neighbourhood instead.
Two things people skip:
- Owner responses are indexable text, and you write them yourself. Answer the question the reviewer implied rather than thanking them.
- Recency counts more than volume. 74% of consumers look for reviews from the last three months.
If volume is the constraint, there are proven ways to generate more Google reviews worth working through. Just write the ask better than most people do.
The third-party slice is small, and it is the part you cannot fix alone
About one citation in eight in my sample came from a page the business didn't own: a Reddit thread, a community blog, a municipal page, and one review platform. Small, but it's also the only slice where writing a better page changes nothing.
That changes what outreach is for. It isn't a volume game against a directory checklist; it's a short, specific list you build by reading your own citation strips.
Run your ten queries, write down every third-party domain that appears, and sort what comes back:
| What showed up | How you get in | Realistic timeline |
| Reddit and forum threads | Answer as a named practitioner; disclose who you are | Ongoing, never promotional |
| Local news and community blogs | Pitch a story, offer a quote, answer a journalist request | Weeks |
| Municipal and association pages | Membership, licensing lists, contractor registries | Weeks |
| Review platforms | Claim the profile, then earn reviews on it | Days to set up |
The top row you can't outreach your way into. The forum threads that got cited were ones where somebody had posted real numbers, which you earn by answering questions under your own name rather than by pitching anybody.
The row where normal outreach applies is local news and community blogs, and there the bottleneck is finding a person. The contact form goes to an inbox nobody has opened since March, and the author byline is a name with no email attached. So I work backwards from the domain instead. Hunter's Domain Search is what I reach for, partly out of habit, and the useful part isn't the addresses so much as the names and job titles next to them.
Run the list through an email verification tool before any of it goes out. I use Hunter's, though anything reputable does the job. Local publisher mastheads go stale within months, and a bouncy first campaign damages the sending domain you'll use all year.
My audit also talked me out of paying for directory listings. Exactly one appeared in sixty citations. If a directory doesn't show up in a citation strip for any of your queries, it isn't part of this project. It may be worth something elsewhere. Not here.
The answer changes depending on where you're standing
Local AI Overviews are generated per location, which breaks the measurement habits everyone carried over from rank tracking.
Checking an overview once from your office tells you what happens on your street. A business can be cited across half a city and absent from the other half, so two people can run the same query, get opposite answers, and both be right.
The method is closer to grid rank tracking than keyword tracking. Take the same ten queries, sample each across a grid of points covering the service area, and at every point record whether an overview appeared, whether you were cited, and which competitors were.
One query, one business, nine sample points, three of them cited. Source: Image by User
Then watch four numbers:
- Trigger rate per query. Some queries quietly stop producing an overview, which hands you the click-back.
- Your citation share, as a percentage of points sampled.
- The same figure for each competitor you care about.
- Which third-party domains keep appearing. This is the list from the previous section, and it's the most useful number here.
One caveat will save you from overreacting. These results are unstable week to week, so read the trend across a month rather than any single check. Track assisted conversions too, since a citation nobody clicks still turns into branded search later.
When a competitor is in the overview and you aren't
Work through it in this order. The likely causes aren't the ones people check first.
- Read the citation strip, not the answer. Note every source the overview pulled for that exact query.
- Check whether the cited pages are competitor sites. In my sample, they usually were. Compare their page to yours on one thing. Does theirs answer the literal question with a number, and does yours?
- If the cited page is a third party, the question is simply whether you're on it. That's an outreach problem, and no amount of on-page work will touch it.
- If you are on the page but still not cited, look at your position and description. I've won placements on roundups that changed nothing for months, always because I sat in the bottom third of the list with a one-line entry that gave nobody anything to quote.
- Compare your review language against the phrasing in the overview. If Google is quoting "same-day service" and not one of your reviews says it, that’s the review script.
- Schema and freshness come last, because they're the least likely cause and the easiest to spend a week on.
The escalation rule: if the same competitor holds three or more of your priority queries, look at what their pages have that yours don't. In my sample, the answer was almost always specificity, not authority.
Turn Local Search Visibility Into a Measurable Growth Plan
Run your own sixty. Take your ten highest-intent research questions, search each one, and write down every domain in every citation strip. It takes about an hour, and it tells you something no general article can: whether your vertical looks like mine, and which specific pages are beating you.
Realistically, the audit is one morning. The pages are a quarter, because ten genuinely specific pages is real writing. Start with the audit, because everything else is guessing until you have it.
Reach out to TechWyse to build a local search strategy that helps your business earn attention before the click, and stay competitive as search continues to change. Call 416-410-7090 or contact us here.
