How SEO changes by industry
A product variant, a model year and a local closing need different answers. Here are the choices I make in six industries, what I check before building a page and the mistake each choice avoids.
The industry decisions below describe my method across my career. Where I can, a short story from my client work shows the approach on a real account, with the client kept anonymous. The numbers all come from my own sites, where I test each principle in public.
E-commerce
A product choice is only useful when fit, stock and delivery answer the same question.
What changes on the page
I keep interchangeable options on one product page. I separate a variant when the search, the specifications and the buying decision are different enough to need their own answer.
Why I make that call: A color name alone is a weak reason to create another page. Compatibility, capacity or a different use can change what the buyer needs to know.
- Who searches
- Shoppers comparing options, often on a phone, often with another tab open.
- What gets built first
- The product and category templates, and the questions people ask right before buying: sizing, shipping, returns, which one to pick.
- The common mistake
- Thousands of thin product pages that all say the same thing, so none of them stands out as the answer.
From my client work, for an e-commerce retailer
Winning the search before the shopper picks a store
I start before the product lands. When new releases are coming, I ask what the market is, who cares about it and who will be searching, and what they'll type before they know which store they want.
Then I match each of those searches to the kind of page that answers it. Some want a product page that answers the buying questions on the page itself. Many want informational content first, written without the store's name in mind, that tells the story around the products so Google and AI answers understand what the store carries and why it's a good place to buy.
I design that route to bring people onto the site through useful information, then connect them to the product pages where they can buy. The goal is the search with no store name in it, from someone who hasn't chosen yet. I want our page to help them make that decision.
Before: the decision is buried
- A separate page for every size
- The same description on each page
- Availability hidden until checkout
After: the answer has a place
One product in three sizes
Size: Medium
- Price
- Price for Medium
- Availability
- Stock for Medium
Size guide and compatibility beside the selection.
Field labels only. No live price, stock or dimensions are shown.
The principle tested on my own sites
Own-site structural example. The variant and buying decisions above are illustrative.
GageRef, my welding and fitting reference, gives each rod and fitting a structured specification page and comparison. The citation chart tracks discovery of those reference pages.
Bing AI citations up 73% in September, a new best month, and the fifth month in a row it grew.
Data through
- May 2026, what else happened: 1 search change
- June 2026, what else happened: 1 search change
- August 2026, what else happened: 1 search change
- September 2026, what else happened: 1 search change
May 2026: what else happened
- Google, : The May 2026 core update. Source: Google Search Status Dashboard; my sites' 14 days either side are on the log.
This is what else was going on that month. I list it for context; nothing here shows any of it moved my clicks.
June 2026: what else happened
- Google, : The June 2026 spam update. Source: Google Search Status Dashboard; my sites' 14 days either side are on the log.
This is what else was going on that month. I list it for context; nothing here shows any of it moved my clicks.
August 2026: what else happened
- Google, : The August 2026 spam update. Source: Google Search Status Dashboard; my sites' 14 days either side are on the log.
This is what else was going on that month. I list it for context; nothing here shows any of it moved my clicks.
September 2026: what else happened
- Google, : The September 2026 spam update. Source: Google Search Status Dashboard; my sites' 14 days either side are on the log.
This is what else was going on that month. I list it for context; nothing here shows any of it moved my clicks.
HI-SCORE the best month so far
| Month | Bing AI citations | Change from the month before |
|---|---|---|
| April | 131 | first month |
| May | 712 | up 444% |
| June | 3.5K | up 386% |
| July | 7.5K | up 115% |
| August | 10.9K | up 46% |
| September | 18.7K | up 73% |
What I got wrong
Welders search by brand name more than I expected.
I had to retrofit brand names and manufacturer equivalents into existing pages after seeing the data.
From my case study, March 1, 2026.
Go deeper: the search moments, the other calls and checks, the method and the tools.
- Search momentsEach moment, and the page that wins it
- Other calls and checks3 situations, and what I check first
- The method5 steps, in order
- The toolsScreaming Frog and Google Search Console
How people search here
Shoppers search in four moments before they buy, and each moment needs a different page.
| The moment | What they type | The page I would build | A search Bing’s AI ran |
|---|---|---|---|
| Working out what kind to buy | how to choose, or a buying guide | A buying guide, or a category page with a short chooser above the grid | under sink water filter buying guide |
| Down to two options | one option vs the other | A comparison page with both side by side in one table | mnpt vs npt |
| Ready for one exact product | a brand, model or part number | The product page, with its specs in a table a machine can quote | e7018 welding rod |
| Looking for the catch | the brand plus problems, reviews or cost | A page that answers the objection plainly, with returns, cost over time and what buyers complain about | PUR water filters cost of ownership criticism reasons |
The last column shows searches Bing recorded, Apr 13 to Oct 11, 2026, while building answers that cited my sites. The fan-reference query is paraphrased to protect its identity; its citation count is unchanged.
Other calls that change by situation
Most e-commerce calls come down to which pages deserve to exist, and the answer changes with the product.
| When | What I do | Why |
|---|---|---|
| A product is out of stock and coming back | Keep the page live and indexed, and say when it's expected back. | It keeps the rankings it earned, and the shopper learns when to return. |
| Filters create thousands of URL combinations | Let search index only the filter pages people search for, like a brand or a size, and keep the rest out. | Crawlers spend their time on pages that can win, and two pages never compete for one search. |
| A product is gone for good | Redirect it to a true replacement. With none, let the page go. | If there is no true equivalent, I would rather return a clean 404 than create a misleading redirect.From my post, February 3, 2026. |
The checks behind the first call
- What I need before publishing
- I compare searches for the product and its variants with the catalog: dimensions, compatibility, stock and price. I check whether variant URLs lead people to the option they searched for.
- What I measure
- I watch which searches reach each page, which options people select and where they leave before adding to cart. Stock changes belong beside that readout.
- What would change the page
- If a variant attracts a distinct need and the shared page cannot answer it clearly, I give it a dedicated page. If several pages answer the same need, I consolidate them.
The method, step by step
Fix the template. One product page template carries thousands of products, so a change to what it shows (specs, sizing, shipping, product photos, reviews) improves every product at once.
Earn trust on the page itself: who is selling, the return policy, reviews from buyers and the details a careful shopper checks before paying. Credibility is what turns a visit into a sale.
Build category pages that help someone choose, with a short answer to which one is right for them above the grid, so the category ranks for comparison searches and not only the product name.
Put the buy link next to the answer, never before it. The shopper who just settled the decision is the one ready to click.
Measure by page type, month over month, so I know which template change moved the numbers.
The tools, and how I use them here
- Screaming Frog
Crawl every product and category page and group the thin, duplicate and broken ones by template, so the fix lands once.
- Google Search Console
Export queries by page type and sort out the questions people ask before buying. I group thousands of them into themes with AI, then decide which themes deserve a page.
Cars
When the model year changes, yesterday's correct specification can become today's wrong answer.
What changes on the page
I separate the current model overview from the year-specific facts. I keep an older year’s useful page and label it clearly, with a route to the current year.
Why I make that call: A used-car shopper still needs the old specifications. Silently replacing them with the new year can answer the search and mislead the buyer at the same time.
- Who searches
- Car shoppers weighing a model over weeks, across several visits and devices, and owners looking for service.
- What gets built first
- The model and trim pages, and the questions shoppers ask in order: which model, which trim, what it really costs, where to buy it.
- The common mistake
- Model pages that repeat the maker's brochure copy, so nothing on them says which trim is worth it.
Before: the decision is buried
- New year added to the old heading
- Old and new trim specs mixed together
- One inventory link for every reader
After: the answer has a place
- Year and trim firstThe heading and comparison table name the same model year.
- Facts tied to that yearEquipment and dimensions sit beside their manufacturer source.
- A useful next stepOlder-year research stays available; current-year shoppers can move on.
The principle tested on my own sites
Own-site structural example; no car account result is shown.
On GageRef, my welding and fitting reference, specifications sit beside their source standard. I bring that source discipline to the illustrative model-year layout.
Go deeper: the search moments, the other calls and checks, the method and the tools.
- Search momentsEach moment, and the page that wins it
- Other calls and checks2 situations, and what I check first
- The method5 steps, in order
- The toolsScreaming Frog and Google Search Console
How people search here
Car shoppers search in stages over weeks, and each stage wants a different page.
| The moment | What they type | The page I would build |
|---|---|---|
| Deciding which model | best, or a size and a budget | A chooser page with a plain pick for each kind of driver |
| Down to two models | one model vs another | A side-by-side table with specs, price range and a short verdict |
| Picking a trim | a model plus trim names, or is it worth it | A trim page that says what each step up adds and what it costs |
| Ready to buy or service | near me, or a model plus inventory or service | A local page with the next step beside the answer |
Other calls that change by situation
Most car calls come down to what happens to a page when the model year turns over.
| When | What I do | Why |
|---|---|---|
| A model is discontinued | Keep the page, say it is discontinued and point to its replacement. | People still search for it, and the replacement page earns the click. |
| Two models could share a comparison | Build it only when people cross-shop them. | A page for every pairing is thousands of near-copies competing with each other. |
The checks behind the first call
- What I need before publishing
- I match the year and trim to manufacturer specifications, then check the page’s search history and the inventory destination. A new badge does not prove that the equipment stayed the same.
- What I measure
- I separate older-year searches from current-year searches and inspect which page receives each. I also check whether inventory clicks land on matching vehicles.
- What would change the page
- If the current overview keeps taking older-year searches, I make the year choices clearer and strengthen the links to the archive. I only retire an old URL after checking its remaining purpose and a relevant destination.
The method, step by step
Build for the order shoppers decide in: which model, then which trim, then what it costs, then where to buy. Each step gets its own page, and each page points to the next.
Put every spec and price in a table with its source and the date it was checked. Model years change, and a stale figure sends the shopper to a page that is no longer right.
Write comparison pages only for the models people really cross-shop, with a plain answer on top about who each one suits.
Put the next step beside the answer, such as local inventory or a service booking, never in front of it.
Measure by page type, month over month, so I know whether model, trim or comparison pages moved the numbers.
The tools, and how I use them here
- Screaming Frog
Crawl the model, trim and inventory pages and group the duplicate and broken ones by template, so one fix covers every model.
- Google Search Console
Export queries by page type and separate the ones that name a model from the ones asking which to buy, since each needs a different page.
Local and multi-location
A closed location still gets searched. The old address needs a clear answer before someone travels.
What changes on the page
I date the closure at the top, verify a relevant alternative and keep the page while people need that answer. I redirect only when the destination fully explains the move.
Why I make that call: A page can still rank after the place has closed. A visit that ends at a locked door is a failure even when the search report shows a click.
- Who searches
- Someone nearby who wants a place, a time or a phone number, usually right now.
- What gets built first
- Two templates done right: a market page for each city and a location page for each place, with the same business details everywhere and the profile that shows in maps.
- The common mistake
- Copy-paste city pages with the town name swapped in. Searchers and Google both see through them.
From my client work, for a multi-market entertainment brand
Fixing pages in batches, not one at a time
For location and market pages, I look for technical changes that can improve them together. I come at it the way a developer would: which batch of changes is worth making across those pages?
I hand those over in batches: a sitemap that lists every page that matters, title tags and metadata written for how people search, schema that tells search engines and AI what each location offers, and a site AI crawlers can read.
Then I tune the location and market pages for the searches that come before anyone picks a brand: date night, family night, something to do this weekend. Those people aren't looking for the brand yet, and the page that answers them is how they find it.
Before: the decision is buried
- Old hours still look current
- Directions point to the closed address
- The closure is buried in a news post
After: the answer has a place
- Status before directionsClosed or moved, with the effective date, above the old address.
- The right alternativeA verified replacement location, its hours and the service it offers.
- One consistent recordThe page, location finder and business profile tell the same story.
The principle tested on my own sites
This is evidence from my own site, in a different field. It shows the method in practice; it is not a result from this industry or a client account.
CheckMyTap, my water quality site, runs on the same two templates: a page for each state and a page for each city, built from the same water data.
From Apr 13, 2026 to Oct 11, 2026, Bing's AI cited 876 CheckMyTap pages. Of the 300 it cited most, 224 are state and city pages. Data through .
What I got wrong
My water-quality catalog shrank in Google: 1.6K pages indexed, then 218 Now a page that breaks the rule doesn't go live. The next step, holding everything when the check itself crashes, is written and not shipped yet. .
Go deeper: the search moments, the other calls and checks, the method and the tools.
- Search momentsEach moment, and the page that wins it
- Other calls and checks2 situations, and what I check first
- The method5 steps, in order
- The toolsGoogle Search Console and Screaming Frog
How people search here
Local searches come in more shapes than near me, and each shape needs a different page.
| The moment | What they type | The page I would build | A search Bing’s AI ran |
|---|---|---|---|
| Right now, nearby | a service plus near me or open now | The map listing first, then the location page it links to | None of my sites serves this moment. |
| Something just happened | the alert or event, sometimes with no place named | A page updated the same day, with the place and the date in its first line | water boil advisory |
| About one place | the city or town plus the question | A city page with something true only of that city | indianapolis drinking water |
| Comparing places | best, cleanest or cheapest, plus a region | One comparison page across every location, built from the same data as the location pages | cleanest water in usa |
The last column shows searches Bing recorded, Apr 13 to Oct 11, 2026, while building answers that cited my sites. The fan-reference query is paraphrased to protect its identity; its citation count is unchanged.
Other calls that change by situation
Every URL should justify being there.
From my post, January 25, 2026.
| When | What I do | Why |
|---|---|---|
| A business that serves an area with no storefront | Pages for the towns really served, each with its own local detail, and no made-up addresses. | A town page with nothing local on it is the copy-paste page from the mistake above. |
| Several locations in one city | A city page that lists them all, and one page for each location. | Searches for the city land on the list, and searches for one place land on its page. |
The checks behind the first call
- What I need before publishing
- I need confirmation from the location owner, the effective date and whether services moved or ended. I check the replacement’s hours, accessibility and service coverage before pointing people there.
- What I measure
- I look at searches for the old location and clicks to calls, directions and the replacement page. I check for stale hours or addresses across the places a visitor might see them.
- What would change the page
- If people still seek the old address, I keep a clear closure answer. If the location moved and a replacement fully answers the same need, I can redirect once the destination explains the move.
The method, step by step
Split the work into two templates: market pages that answer what's near me in this city, and location pages that answer what's at this one place, when, and how to get there.
Give every location page something only that location has: hours, parking, local details and photos of the place. A page with nothing unique doesn't get built.
Keep the name, address, phone and hours identical across the website, the map profiles and the directories, checked on a schedule instead of when someone notices.
Improve the template once and every market and location page gets better together, instead of fixing pages one at a time.
Measure by market and by location, so one strong city can't hide five weak ones.
The tools, and how I use them here
- Google Search Console
Filtered by the location folder and by city searches, so each market gets its own trend line.
- Screaming Frog
Every location page crawled for missing hours, duplicate titles and pages no menu links to.
Legal
A legal answer without its jurisdiction and a qualified review can mislead the person who needs it.
What changes on the page
I make the jurisdiction and review date visible before the answer. A separate page earns its place when a different rule changes what the reader needs to do.
Why I make that call: Swapping a state name into the same article can hide a meaningful difference. A clear answer still needs the right scope and qualified review.
- Who searches
- Someone worried, sometimes in a hard moment, looking for a clear answer and someone to trust.
- What gets built first
- Plain-language answers to the questions people actually ask, reviewed for accuracy, with the credentials behind them visible.
- The common mistake
- Writing for other lawyers instead of for the person who's scared and searching at night.
From my client work, for an AI software company serving the legal industry
Writing for the reader who hasn't heard of the product
I write for people looking for help with a problem before they know the brand.
So the blog is written for them: articles aimed at the businesses and teams who would actually use the service, each one answering the problem they searched for and then pointing to the solution and platform pages that solve it.
The work is getting someone from a question to the other side, from not knowing the brand to looking at the product that fits.
Before: the decision is buried
- A broad answer with no jurisdiction
- A date that only says the page changed
- A form before the limits of the answer
After: the answer has a place
- Where the answer appliesThe jurisdiction and question appear together in the heading.
- The reviewed explanationThe relevant authority, effective date and reviewer sit with the answer.
- Limits before the next stepExceptions and missing facts explain when individual advice is needed.
The principle tested on my own sites
Own-site source and correction example. The legal review process above is an illustrative standard.
WireRef shows source-linked electrical code values and the public corrections they triggered. Its click chart measures discovery.
Search clicks held level in September, up 2% on August, a new best month by a hair.
Data through
- May 2026, what else happened: 1 search change
- June 2026, what else happened: 1 search change
- August 2026, what else happened: 1 search change
- September 2026, what else happened: 1 search change
May 2026: what else happened
- Google, : The May 2026 core update. Source: Google Search Status Dashboard; my sites' 14 days either side are on the log.
This is what else was going on that month. I list it for context; nothing here shows any of it moved my clicks.
June 2026: what else happened
- Google, : The June 2026 spam update. Source: Google Search Status Dashboard; my sites' 14 days either side are on the log.
This is what else was going on that month. I list it for context; nothing here shows any of it moved my clicks.
August 2026: what else happened
- Google, : The August 2026 spam update. Source: Google Search Status Dashboard; my sites' 14 days either side are on the log.
This is what else was going on that month. I list it for context; nothing here shows any of it moved my clicks.
September 2026: what else happened
- Google, : The September 2026 spam update. Source: Google Search Status Dashboard; my sites' 14 days either side are on the log.
This is what else was going on that month. I list it for context; nothing here shows any of it moved my clicks.
HI-SCORE the best month so far
| Month | search clicks | Change from the month before |
|---|---|---|
| April | 655 | first month |
| May | 1.1K | up 69% |
| June | 1.3K | up 21% |
| July | 1.7K | up 31% |
| August | 2.8K | up 59% |
| September | 2.8K | up 2% |
What I got wrong
A WireRef page said a wire carried 200 amps. It carries 180. A check now compares every ampacity claim on the site with the code table. It was proven to fail on the old pages before it shipped. .
Go deeper: the search moments, the other calls and checks, the method and the tools.
- Search momentsEach moment, and the page that wins it
- Other calls and checks2 situations, and what I check first
- The method4 steps, in order
- The toolsGoogle Search Console and Peec AI
How people search here
People with a legal question search four ways, and every one of them needs the plain answer first.
| The moment | What they type | The page I would build | A search Bing’s AI ran |
|---|---|---|---|
| Something just happened to them | what to do after, can they, is it legal to | A plain-language answer page with the short answer, the steps, then when to get help | None of my sites serves this moment. |
| Their own situation | the question with their own details in it | An explainer with a table where they find their row | wire size for 60 amp breaker |
| Weighing two options | one option vs the other | A comparison that says which one fits whom | thhn vs xhhw |
| Looking up the rule itself | the code, statute or article number | A page per rule with the text, what it means in plain words and the date it was checked | nec article 230 |
The last column shows searches Bing recorded, Apr 13 to Oct 11, 2026, while building answers that cited my sites. The fan-reference query is paraphrased to protect its identity; its citation count is unchanged. My real searches here come from WireRef, my electrical code site, another field where a wrong answer can hurt someone.
Other calls that change by situation
In legal, the hard calls are about how much to say and when to send someone to a lawyer.
| When | What I do | Why |
|---|---|---|
| A rule changes | Update the page, change its date and note on the page what changed. | The reader can see the answer is current, and so can anyone checking it. |
| A question is too risky to answer in general | Answer what's true for everyone, then say plainly when to talk to a lawyer. | The reader still leaves with a next step, and the page never gives advice it can't stand behind. |
The checks behind the first call
- What I need before publishing
- I need the primary authority, its effective date and a qualified reviewer’s confirmation that the explanation fits this jurisdiction. An automated content refresh cannot stand in for that review.
- What I measure
- I check which jurisdiction-specific searches land here and whether the page sends readers to the right next step. Traffic alone cannot establish that a legal explanation is correct.
- What would change the page
- If the rule changes, I flag the affected answer for review and update the explanation, source and date together. If two pages no longer differ in a useful way, I combine the shared answer and retain the real distinctions.
The method, step by step
Collect the questions people really ask, in their own words, from search data, People Also Ask and the forums where they ask each other.
Answer each one in plain language in the first few sentences, then go deeper for the reader who needs more.
Give every claim a source and a review by a qualified professional before it ships, and show who reviewed it on the page.
Re-check pages on a schedule, because rules change and a stale answer is a wrong answer.
The tools, and how I use them here
- Google Search Console
Question-shaped searches grouped by topic. I group them with AI, then decide which questions deserve their own page.
- Peec AI
Which sources AI answers cite for these questions, and whether someone else's explainer is the one being quoted.
Real estate and housing
A buyer's question changes by place. I separate the local demand before choosing a page or judging results.
What changes on the page
I fold it into a regional page until I can explain what changes in that market: available homes, costs, timing or constraints that affect the decision.
Why I make that call: A city name and a few familiar landmarks do not make a useful market page. The reader needs to know what living or buying there changes for them.
- Who searches
- People making a big, slow decision about where to live, comparing places and prices over months.
- What gets built first
- A map of the themes people search, with branded and non-branded searches kept apart.
- The common mistake
- Reporting one blended number. Branded searches hide how you're doing with everyone who hasn't heard of you yet.
From my client work, for a national homebuilder
Meeting the move before the house search
For searches without a builder's name, I start with the place: is this state, this city, this part of town right for us?
So I build out the state, city and regional pages to answer that, with real value for someone thinking about moving there, before they've picked a builder. Each page then connects to the homes for sale in that area.
I design that route to help someone move from a question about a place to the homes listed there.
Before: the decision is buried
- City name pasted into a general guide
- Undated prices with no source
- A contact form as the only next step
After: the answer has a place
- Start with the regionA useful regional answer holds markets that lack distinct evidence.
- Show the local differenceDated inventory or cost context appears only where the source supports it.
- Make comparison possibleReaders can compare nearby options and see what remains unknown.
The principle tested on my own sites
Both examples come from my own sites. No housing-market or client result is shown.
My public dashboard separates Google and Bing clicks by site and month. RatingPlate’s guide received more Bing AI citations than its 25 cited state guides combined.
RatingPlate has 51 state cost pages. In Bing's AI page export through Oct 11, 2026, its one guide to reading a water heater's label drew 58.5% of the citations, and the 25 state pages it cited drew 7.4% between them. Data through .
Search clicks up 21% in September, a new best month, and the fifth month in a row it grew.
Data through
- May 2026, what else happened: 1 search change
- June 2026, what else happened: 1 search change
- August 2026, what else happened: 1 search change
- September 2026, what else happened: 1 search change
May 2026: what else happened
- Google, : The May 2026 core update. Source: Google Search Status Dashboard; my sites' 14 days either side are on the log.
This is what else was going on that month. I list it for context; nothing here shows any of it moved my clicks.
June 2026: what else happened
- Google, : The June 2026 spam update. Source: Google Search Status Dashboard; my sites' 14 days either side are on the log.
This is what else was going on that month. I list it for context; nothing here shows any of it moved my clicks.
August 2026: what else happened
- Google, : The August 2026 spam update. Source: Google Search Status Dashboard; my sites' 14 days either side are on the log.
This is what else was going on that month. I list it for context; nothing here shows any of it moved my clicks.
September 2026: what else happened
- Google, : The September 2026 spam update. Source: Google Search Status Dashboard; my sites' 14 days either side are on the log.
This is what else was going on that month. I list it for context; nothing here shows any of it moved my clicks.
HI-SCORE the best month so far
| Month | search clicks | Change from the month before |
|---|---|---|
| April | 1.7K | first month |
| May | 4.4K | up 159% |
| June | 7.6K | up 74% |
| July | 14.2K | up 87% |
| August | 19.7K | up 39% |
| September | 24K | up 21% |
Go deeper: the search moments, the other calls and checks, the method and the tools.
- Search momentsEach moment, and the page that wins it
- Other calls and checks2 situations, and what I check first
- The method5 steps, in order
- The toolsPeec AI and Google Search Console
How people search here
Housing decisions take months, and the searches narrow from a whole region down to one exact home.
| The moment | What they type | The page I would build | A search Bing’s AI ran |
|---|---|---|---|
| Deciding where | a place plus cost, schools, commute or what it's like to live there | An area guide with real local data and the date it was read | None of my sites serves this moment. |
| Working out what fits | sizes, options and what's available | One page that lays every option out in a table | heat pump water heater sizes available |
| Down to one exact one | an address, a listing or a model number | The page for that one, with every detail written out | rheem proe50 t2 rh95 spec sheet |
The last column shows searches Bing recorded, Apr 13 to Oct 11, 2026, while building answers that cited my sites. The fan-reference query is paraphrased to protect its identity; its citation count is unchanged. My real searches here come from RatingPlate, my heat pump water heater site, another big and slow purchase for a home.
Other calls that change by situation
In housing, the calls are about where to split a page by place and where to keep one page for everyone.
| When | What I do | Why |
|---|---|---|
| Prices or rules change by place | One page with a table by place, dated. | One page gathers the searches, and the date tells the reader how fresh it is. |
| A question everyone has, wherever they live | One page that answers it for every market, linked from each market page. | It gathers every market's searches in one place. On RatingPlate, that one page drew more AI citations than the state pages together. |
The checks behind the first call
- What I need before publishing
- I check the geography and date behind each market claim, the available inventory and whether the local differences matter to the buyer. Small or stale samples stay labeled.
- What I measure
- I separate searches for the region from searches for each market, then look at movement into listings or comparison pages. A changing inventory can explain changing behavior.
- What would change the page
- When a market has sustained demand and enough distinct evidence, I give it a page. If that evidence disappears, I narrow the claims or return the answer to the regional guide.
The method, step by step
Split demand into the few themes people actually search, and give each theme its own pages and its own measures.
Measure each theme by market, because the same question means something different from one town to the next.
Separate branded from non-branded searches in every report. Non-branded is where new buyers find you, and where the growth is.
Write the questions a buyer would ask an AI assistant about each theme, to see who gets recommended and why.
Find the questions nobody answers well, and build the content only you can write.
The tools, and how I use them here
- Peec AI
Share of voice, citation share and sentiment for each theme's question set.
- Google Search Console
A branded and non-branded split with a filter, then each theme's searches grouped by market.
Publishing and reference sites
A reference page should answer exactly what is known, show its source and say what remains unknown.
What changes on the page
I lead with what is confirmed and say plainly what is still unknown. A rumor does not become a release date because people are searching for one.
Why I make that call: The reader needs the current answer, even when that answer is that nobody has announced it yet. A confident guess can keep circulating after the page changes.
- Who searches
- Fans and professionals who want one exact answer, fast, and will come back if it's right.
- What gets built first
- The single question each page answers, kept current, with the sources visible.
- The common mistake
- Publishing fast and fixing later, on pages people come to for one exact answer.
Before: the decision is buried
- A predicted date in the headline
- Rumor repeated as fact
- No visible record of what changed
After: the answer has a place
- Answer the question firstNo confirmed release date, if that is what the evidence supports.
- Separate known from unknownThe announcement source and last check sit beside the confirmed facts.
- Update the same answerA change note explains a confirmed announcement when it arrives.
The principle tested on my own sites
This is a public example from my own publishing site. Its measured results and corrections are in the case study.
I built a release-answer page for an independent fan reference before there was a confirmed date.
Google clicks up 17% in October, a new best month, and the sixth month in a row it grew.
Data through
- May 2026, what else happened: 1 search change
- June 2026, what else happened: 1 search change
- August 2026, what else happened: 1 search change
- September 2026, what else happened: 1 search change
- October 2026, what else happened: 1 search change
May 2026: what else happened
- Google, : The May 2026 core update. Source: Google Search Status Dashboard; my sites' 14 days either side are on the log.
This is what else was going on that month. I list it for context; nothing here shows any of it moved my clicks.
June 2026: what else happened
- Google, : The June 2026 spam update. Source: Google Search Status Dashboard; my sites' 14 days either side are on the log.
This is what else was going on that month. I list it for context; nothing here shows any of it moved my clicks.
August 2026: what else happened
- Google, : The August 2026 spam update. Source: Google Search Status Dashboard; my sites' 14 days either side are on the log.
This is what else was going on that month. I list it for context; nothing here shows any of it moved my clicks.
September 2026: what else happened
- Google, : The September 2026 spam update. Source: Google Search Status Dashboard; my sites' 14 days either side are on the log.
This is what else was going on that month. I list it for context; nothing here shows any of it moved my clicks.
October 2026: what else happened
- Google, : More links in AI Mode and AI Overviews. Source: The Keyword, Google's blog; my sites' 14 days either side are on the log.
This is what else was going on that month. I list it for context; nothing here shows any of it moved my clicks.
HI-SCORE the best month so far
| Month | Google clicks | Change from the month before |
|---|---|---|
| April | 42 | first month |
| May | 491 | up 1,069% |
| June | 1.2K | up 153% |
| July | 5.7K | up 357% |
| August | 9.3K | up 64% |
| September | 13.8K | up 47% |
| October | 16.2K | up 17% |
What I got wrong
I made a release-date answer sound more certain than its sources A tighter check around the page. A dated page now has to stay as careful as its sources. .
Go deeper: the search moments, the other calls and checks, the method and the tools.
- Search momentsEach moment, and the page that wins it
- Other calls and checks2 situations, and what I check first
- The method4 steps, in order
- The toolsGoogle Search Console and Bing Webmaster Tools
How people search here
Reference searches come in bursts around news and seasons, and each burst wants one exact answer.
| The moment | What they type | The page I would build | A search Bing’s AI ran |
|---|---|---|---|
| Waiting on news | when is, release date, next | One page per big question, with what's confirmed, what's rumored and the date it was checked | When is the next adaptation coming?(query paraphrased to protect project identity) |
| During the season | roster, depth chart, injuries or schedule, plus the year | A page per team question, updated the day it changes | oregon depth chart 2026 |
| Looking up a fact | a name, a term or a number | A page per person or thing, with the facts in a table and their sources | oregon coaching staff |
The last column shows searches Bing recorded, Apr 13 to Oct 11, 2026, while building answers that cited my sites. The fan-reference query is paraphrased to protect its identity; its citation count is unchanged.
Other calls that change by situation
Plausible and accurate are not the same thing, and on reference sites, the gap between the two has real consequences if not done right.
From my post, March 4, 2026.
| When | What I do | Why |
|---|---|---|
| Two pages could answer one question | Keep one page and put the second angle inside it. | Two pages split the same searches, and neither becomes the answer. |
| A page's question has passed | Keep it, and update it with what happened. | The same question comes back next season, and the page is already there. |
The checks behind the first call
- What I need before publishing
- I check the original announcement or the creator’s own update. I distinguish what the source actually says from a fan’s interpretation, and record when I checked it.
- What I measure
- I look at the release-date searches reaching the page and whether readers find the source or latest update. More traffic does not turn speculation into confirmation.
- What would change the page
- When a primary source confirms the date, I update the headline, answer and change note together. If a claim is withdrawn, I correct the page visibly rather than quietly replacing it.
The method, step by step
Pick the one question each page answers before a word gets written.
Use primary sources only, and show them on the page with their dates.
Build the page for a question before it peaks, like a release date nobody has announced yet.
Keep pages current on a schedule, and log every correction in public.
The tools, and how I use them here
- Google Search Console
Daily exports to spot a page climbing before it peaks.
- Bing Webmaster Tools
Bing clicks and AI citations, which often move first on a new reference site.
The stakes change the page
Old car specifications can mislead a buyer. Wrong opening hours waste a trip. An unreviewed legal answer can give false confidence. I change the page, the evidence and the review process to fit the decision.
These are three tests of the same experience. The decisions above show what changes when the reader's situation changes.
The toolkit, and the question each tool answers
Use Peec for answers, Bing for citations, GSC for phrasing, and GA4 for the money end, and run the loop: listen, mine, answer, verify.
That's from my post on how AI picks which brands to name, published July 18, 2026. Where each tool sits in my system: the outcome it serves (get found, get understood, get chosen) and the pillars of how I work. The full map is in my SEO system.
Peec AI
How often do AI answers mention us, which pages do they cite, and how do they describe us?
A fixed set of prompts for each theme, run on a schedule: share of voice, citation share and sentiment, with competitors side by side.
Serves: Get found, Get understood. Pillars: Be the source worth citing, Measure, then double down.
Ahrefs Brand Radar
Where does the brand get cited and mentioned across AI answers, compared with competitors?
Citation and share of voice trends across a bigger set of prompts than I could track by hand, built from the questions I wrote first.
Serves: Get found, Get chosen. Pillars: Be the source worth citing, Measure, then double down.
Bing Webmaster Tools
What does first-party data say about AI citations and Bing search?
Bing's AI performance report counts citations in Bing's AI answers, and Bing clicks often move before Google's on a new site.
Serves: Get found. Pillars: Never ship a wrong fact, Measure, then double down.
Google Search Console
What do people actually search, and which pages are they finding?
Bulk exports with far more rows than the interface shows. AI helps me sort a mountain of queries into themes per pillar, which would take weeks by hand.
Serves: Get found, Get understood. Pillars: Start from the person, Measure, then double down.
Google Analytics 4
Did the visit turn into the action that matters?
Key events and revenue by landing page and channel, including visits from AI assistants.
Serves: Get chosen. Pillars: Measure, then double down.
Screaming Frog
Can a search engine reach and read every page?
Full crawls for broken links, redirect chains, missing tags, duplicate titles and orphan pages, grouped by template.
Serves: Get found. Pillars: Never ship a wrong fact, Turn every lesson into a system.
Technical audits, by hand and with AI
What's actually broken, and what matters most?
A crawl, the indexing reports and a manual walk through the site on a phone, the way a reader sees it. AI sorts the thousands of rows; the walk-through catches what no export shows.
Serves: Get found, Get understood. Pillars: Never ship a wrong fact, Turn every lesson into a system.
Microsoft Clarity
What do people do once they're on the page?
Heatmaps and session recordings on the pages that matter most.
Serves: Get understood, Get chosen. Pillars: Start from the person, Tell it as a story.
What never changes
What stays the same in every industry: start from the person and the question they typed, build the page that answers it best, and measure whether they found it, understood it and chose it.
The same method on my own sites, month by month or back to how I work at my full-time job. I've also written about industry SEO.