Four SEO Checks We Didn’t Find in Other Audit Tools

Run almost any SEO audit tool and you get a familiar list: broken links, missing titles, redirect chains, slow pages. Those problems are real, and DeepSpider.ai checks them too. But some of the problems we keep seeing on real websites only show up when you compare one thing with another: a title with the page under it, a link with the page it leads to, or one page with fifty others that look almost the same.

We compared DeepSpider’s checks with the published check lists of 27 audit tools, including Screaming Frog, Sitebulb, Lumar, Ahrefs and Semrush. The four checks below didn’t appear in any of those lists in this form. A tool can have a check without documenting it, so read that as “not found in the tools we reviewed”, not as a promise that nobody has one.

In short
These four checks look at problems that sit between pages: a “best” list that puts its own product first, a title or description that promises a year, discount or price the page doesn’t show, link text that names a different page on the same site, and location pages that repeat one text with only the place name changed. None of them changes Website Health Score. Each one points you to pages that deserve a second look.

1. “Best” lists that put your own product first

A project management company publishes “12 Best Project Management Tools in 2026” on its blog. Number one is its own product. The other eleven are competitors, each with a short paragraph. To someone arriving from search it looks like an independent comparison, but it’s closer to an advert.

The tactic spread because these lists are what people search for and what AI assistants quote. In February 2026 Lily Ray reported that SaaS and B2B blogs with dozens or hundreds of such lists lost 29 to 49% of their Google visibility after January. She was clear that this is an association, not a proven cause. Devesh Khanal of Grow and Convert disputes the link: on the sites he looked at, the lists were under 10% of the posts, and he sees weak content in general as the likelier cause.

The AI side is mixed too. Peec analysed 232,000 citations in the software niche and found that about 11% came from lists where the publisher ranked itself first, with no sign of that falling over 12 weeks. In October Ray ran 100 business software queries: when Google’s AI answers cited a self-ranked list, the brand behind it was left out of the recommendation 83% of the time, PPC Land reports. It’s a small test in one niche, but it suggests the list can get cited while the product it was written for doesn’t.

Google hasn’t announced anything about these lists. Its guidance on reviews asks that when you call something the best, you explain why, with first-hand evidence.

What the check looks for

DeepSpider reports a page when its title or H1 is a “best” or “top” list with a number or a year, the main content has at least three list entries as headings, the first entry names the website itself and at least two other entries don’t. The website’s name comes from the site name repeated in its page titles, its WebSite or Organization structured data, or its domain. Titles in English, German, Estonian, Spanish, Portuguese, Italian, French and Japanese are read.

It leaves out lists about the company itself (“Best HubSpot alternatives” on HubSpot’s own blog), lists of the company’s own products, and advice lists such as “10 best practices”. On 11 live comparison pages it flagged the three that put their publisher first and none of the others. It also missed one, where the products sat in cards rather than headings.

What to do

• Say at the top, in plain words, that the list is written by the company behind the first product.
• Explain how you chose and tested the products, and give each place its reasons and evidence.
• If you can’t show that, compare by use case (“best for small teams”) instead of ranking yourself first.
• Don’t change only the year in the title. Change it when you’ve actually re-checked the entries.

2. Titles and descriptions that promise what the page doesn’t show

A title says “Spring sale: 30% off all bikes”. The sale ended in May and the page no longer mentions a discount. Or a guide’s title still says 2026, the text hasn’t been touched since 2024, and the only “2026” on the page is in the copyright line.

Google builds snippets mainly from the page content and rewrites titles that don’t describe the page. Its title link documentation names this exact case: a page used year after year whose title wasn’t updated, where Google may take the year from a visible heading instead. In his Zyppy study Cyrus Shepard found Google rewrote 61.6% of 81,000 titles. That measures what Google shows, not rankings. The bigger cost is trust: someone who clicks for 30% off and doesn’t find it leaves.

What the check looks for

DeepSpider takes the years, percentages and currency amounts in each title and meta description and looks for them in the page text, including headers, footers and image alt text. It reads numbers the way people write them, so “1.234”, “1,234” and “1234” all match, and a price shown in another currency still counts. Copyright years don’t count as found. Counts such as “15 best”, bounds such as “under $300”, ranges and per-unit prices aren’t treated as claims, because small numbers turn up on almost any page by chance.

On a 300-page recording of one large Japanese site it found 62 pages whose title said 2026 while the page showed the year only in its copyright line. On 17 other sites it found nothing, which is what you’d hope: most titles match their pages.

The check says a claim wasn’t found, not that it’s false. A price loaded later by JavaScript isn’t in the HTML DeepSpider reads, for example. So open the page before changing anything. If the offer still stands, show it as text; if it changed or expired, update the title and description. If the page is meant to be current, update the content first and the year second.

The closest check we found elsewhere is Ahrefs’ “Page and SERP titles do not match”, which compares your title with what Google shows. That’s useful, but it doesn’t tell you whether the page itself backs up the title.

3. Link text that names a different page

On a Japanese casino guide, a table of game types linked the word “Roulette” to the site’s minimum deposit page. The site had a roulette page, and two other pages linked “Roulette” to it correctly. One table had a copy and paste slip, easy to make and almost impossible to spot by reading the HTML.

Google asks for anchor text that is “descriptive, reasonably concise, and relevant to the page that it’s on and to the page it links to”. Nielsen Norman Group calls a link a promise the destination should keep. John Mueller has said that changing internal anchor text isn’t something you’d see a visible effect from in search, so treat this as a clarity problem. Visitors land somewhere they didn’t expect, and the page the words describe loses an internal link that clearly names it.

What the check looks for

Plain word matching would flag every synonym and menu label, so this check is strict on purpose. A link is reported only when its text shares no word with the title, H1 or address of the page it leads to, another audited page has every word of the text in its title or H1, and the site itself uses the same text for that other page on at least two other pages. Generic labels such as “Read more”, “Contact” or “Terms”, links to the homepage and texts that name the site are left out.

That makes findings rare by design. On 20 sites we recorded it found nothing; on a longer recording of the casino guide it found the roulette link. When it does report something, it’s usually a real mistake. Fix it by pointing the link at the page its text describes, or by changing the text to describe the page it leads to. If the link sits in a menu or template, fix it once there.

Other tools list the same text pointing to several pages (DeepSpider has that view too), and JetOctopus, Botify and InLinks let you review anchor text by hand. We didn’t find one that checks automatically whether the words match the destination.

4. Location pages built from one template

A transcription company has a page for every US state: “Texas Transcription Services”, “Ohio Transcription Services” and so on. Each page says the same thing with the state name swapped in. Each state page also lists dozens of cities as links, but most of those links redirect straight back to the state page.

That isn’t a made-up example. On one site we tested, DeepSpider grouped 147 such pages whose main text matched 98 to 100%. Their state pages linked to 4,628 city URLs, and only 100 of them were real pages.

Google’s spam policies name both patterns under doorway abuse: “pages targeted at specific regions or cities that funnel users to one page” and “substantially similar pages that are closer to search results than a clearly defined, browseable hierarchy”. Practitioners agree on the cure. Despina Gavoyannis at Ahrefs and Melissa Popp at BrightLocal both warn against copying one page and swapping the city name, and Miriam Ellis at Whitespark notes that shared sections such as a service list can repeat as long as the page adds something local. Sterling Sky offers a fair counterpoint: similar service area pages can still bring traffic, and thin pages are the bigger risk.

What the check looks for

Most audit tools flag near-duplicate pages. This check looks for a specific pattern: at least 10 indexable pages whose titles follow one template with one to four words changed, where those words are places (states, provinces, countries, “City, ST” names or a word after “in” or “near”), and whose main text is at least 80% the same once the place names are removed. Language versions linked by hreflang, noindex pages and groups where most pages carry their own street address in structured data (real branches) are left out. A legal site’s 15 state law pages, each written for its state, weren’t reported.

A related change sits in Internal links to redirects: links whose target redirects back to the page that links to it are now marked “redirects back to this page”. On the transcription site, that’s the list of city links that go nowhere.

What to do

• Check Search Console first. Keep the pages that bring clicks and give them something local: customers, projects, people, rules or prices from that place.
• Merge the rest into one service area page and redirect the copies in batches, comparing clicks before and after.
• Show places you only serve as plain text, not as links that send visitors back where they started.

How DeepSpider.ai fits in

All four checks run on every new audit and use evidence the crawl already saves, so they add no requests to your website. None of them changes Website Health Score, because each can match a page for a good reason: an honest comparison, a price shown by JavaScript, a deliberate link or a branch with its own page. Two are Needs review, where a person has to decide, and two are Informational. Each finding comes with How to fix it guidance that picks the advice matching what the audit found, with its sources. Audits run before these checks existed show them as Added later; run a new audit to see them.

The checks compare words and numbers, not meaning. DeepSpider reads the HTML your server sends and the first 4,000 characters of each page’s main text, so it can miss text added later by scripts. If you want more on keeping old pages honest, our guides to content decay and AI-generated content go into more detail.

Frequently asked questions

Is ranking your own product first against Google’s rules?

Google hasn’t said so, and its spam policies don’t mention self-ranked lists. Its reviews guidance asks you to explain why something is the best, with first-hand evidence. The risk is that the list reads as independent when it isn’t, which costs you trust with readers and possibly with AI answers.

Should I delete all my city pages?

No. Start with Search Console clicks, keep and improve the pages that bring customers, and merge the rest gradually. If you delete everything at once, you can’t tell what helped.

The short version

These checks compare things that each look fine on their own. A list that crowns its author, a title that promises more than the page, a link to the wrong page and a set of copied city pages can all pass a standard audit. Look at what DeepSpider finds, decide page by page, and fix the content before the title or the year.

Sources