Faragiri

Learn On page SEO · Title tags, Meta descriptions, Headings and 4 more

One rewritten title at a Shrewsbury bee supplier, the position unmoved and the clicks more than doubled, and the six other things a page controls.

7Branches
7Nodes below
3Depth

Thornfield Bees in Shrewsbury had a guide called Autumn Feeding | Thornfield Bees. On 3 November 2025 the title tag was changed to When to stop feeding bees in autumn: Shropshire dates. The page held position 6.1 in October and 6.0 in December, and clicks went from 210 a month to 480 while the position held at six.

That is the largest effect a single line of markup has available to it, and it is the whole argument for treating the title tag as the one thing on a page worth an afternoon. Nothing about the page's standing changed. What changed is that the line in the results now contained the words somebody had typed and then the answer to them.

The rule the rewrite followed

Put the words somebody typed at the front of the title, and the answer after them. Autumn Feeding is a category name and it answers nothing; when to stop feeding bees in autumn is a sentence a person types with a hive in front of them, and Shropshire dates promises the specific thing that ends the search.

Title, h1, and description

How the three listing lines divide the work on a page

Put typed search words and the specific promise in the title tag, keep the h1 as a reader safety net, and spend twenty seconds on the description.

Title tagLives in the page head for the search listing
Meta descriptionSummarises the listing for the person choosing
Effect on positionDirectly shapes how the page matches the queryNone; affects clicks rather than ranking
Rewriting ruleLead with typed words, follow with the specific answerState the answer plainly so search snippets keep it
When the engine swaps itFalls back to the on-page h1 if the title misses the queryDiscards it for a matching sentence lifted from the page
Time worth spendingAn afternoon on pages that matterTwenty seconds, and no more than that

A strong h1 is a safety net for a weak title tag, never the other way round.

Two things about the title are worth having straight before rewriting one. The title tag is not the h1, and the engine may replace either one: the h1 sits in the page for a reader, the title tag sits in the head for the listing, and when the title does not match the query well, a heading from the page gets used instead. So a good h1 is a safety net for a bad title and never the other way round.

The description is the other half of the listing and it does none of the same work. A meta description does not affect the position, only the click, and it is discarded whenever it does not answer the query that was typed, with a sentence lifted out of the page put in its place. Writing one is worth twenty seconds and worth no more than that.

What the page is actually made of, from the bottom

Under the title, the guide answers with a date: the first frost in Shropshire in 2025 was on 28 October, and the syrup comes down when the bees stop taking it rather than on a calendar day. That paragraph is the base of everything else on this page. A page that ends the search can carry a mediocre title. A superb title on a page that does not answer buys one click and gives it back.

On the page

Five on-page decisions, drawn by how much has to sit underneath them

The tip is worth doing and the base is worth existing for, and almost every audit spends its time at the tip because the tip is countable.

  1. Structured dataChanges the shape of the listing. Never the position.
  2. Alt text and filenamesWritten for a person who cannot see the photograph.
  3. The title tagThe one line with a number attached to it.
  4. Headings and internal linksThe outline, and the claims this page makes about others.
  5. A page that ends the searchHolds all four above it up, and no crawler can measure it.

Volume, not order. Nothing here happens before anything else; the base is simply the one the other four are worthless without.

Above that sits the outline. Headings are the shape of the answer, and a reader skimming and a crawler parsing use the same one. Above that, the title. Above that, alt text is written for somebody who cannot see the image, and the filename is the smaller half of the same job: the photograph on this page is a full super and the file is called autumn-feeder-thornfield.jpg, which helps a little and helps less than the sentence describing it.

At the tip, structured data changes what the listing looks like, not where it sits. It is worth doing and it is worth knowing what it buys, and what it buys is a different-shaped result rather than a higher one.

The link out is a claim about somewhere else

The anchor text on a linking page is a claim about the page it points at, which makes internal linking the one on-page decision that mostly changes somebody else's page. Thornfield's guide links to a page on moving hives in winter using those words rather than the page's own name, and the linked page collects that description from every page that points at it.

Internal anchor text

A link out is a claim made on behalf of the destination page

Using plain descriptive words tells the target page what it covers; repeating the exact same twelve-word phrase forty times adds nothing.

Linking guideAutumn feeding page referencing winter hive care
  1. 01Descriptive wordingUses 'moving hives in winter' rather than a page title string
  2. 02Distinct facetsDifferent source pages describe different angles of the topic
  3. 03Avoids identical repeatsForty identical claims collapse into one claim repeated forty times
Linked destinationCollects descriptions from every page pointing at it

Internal linking is the one on-page choice that mostly changes what another page means to search engines.

Which is why the same twelve words used as anchor text on forty pages does nothing. Forty identical claims are one claim, made forty times, and the pages that could have described a different facet all said the same thing instead.

Where this stalls: auditing what is easy to count

The stall is that six of the seven things on this page can be counted by a machine and the seventh cannot. Title length, heading order, missing alt text and absent structured data all appear on a report as a number that can be driven to zero, and driving it to zero is satisfying, cheap and mostly worth nothing.

Where audits stall

Six items a crawler can count to zero, and the one that decides

Thornfield spent a fortnight in 2025 fixing 400 alt attributes and moved nowhere because the page title still failed to answer the search.

  • 400 missing alt attributesEasy to count and satisfying to drive to zero
  • Title length and heading orderFlagged mechanically in every crawl table
  • Absent structured dataReported as errors regardless of page quality
What automated audits report
  • Whether the page ends the searchDecides the outcome, yet no crawler can measure it
  • Whether the title matches typed queriesAutumn Feeding hid the real Shropshire dates question

Driving countable crawl warnings to zero is cheap and reassuring, and mostly worth nothing when the answer itself is wrong.

Whether the page ends the search is the one that decides the outcome, and no crawler reports it. Thornfield spent a fortnight in 2025 fixing 400 alt attributes and the guide moved nowhere, because the guide had not been short of alt text. It had been called Autumn Feeding.

What to open in 2026

Screaming Frog SEO Spider reads every title, description and heading on a site in one pass, free up to 500 addresses, which covers Thornfield four times over. It replaces opening each page and reading the source by hand, and its most useful column is the one listing pages whose title is duplicated somewhere else.

What to open in 2026

Two free tools read the markup; only a person checks the fact

Screaming Frog audits titles and headings across 500 URLs, Rich Results Test checks schema, and neither knows when Shropshire frost arrived.

  • Screaming Frog SEO SpiderFree up to 500 addresses
    • Reads every title, description, and heading in one pass
    • Finds titles duplicated across multiple pages
  • Domain verificationNo tool checks truth
    • Crawler sees a date inside a heading
    • Only a Shropshire beekeeper knows 28 October was right
2026 page checks
  • Rich Results TestReplaced Google's 2020 tool
    • Confirms structured data is valid
    • Shows which listing shape the markup can earn

Watch out for old tutorials pointing to the retired Structured Data Testing Tool handed to schema.org in 2020.

The Rich Results Test checks whether structured data is valid and what it would earn. It is free, and it replaced the Structured Data Testing Tool that Google retired in 2020 and handed to schema.org, so a tutorial pointing at the old one is pointing at a different organisation's product.

Neither one can say whether the answer is right. A crawler confirms the date is in a heading; only a beekeeper in Shropshire knows whether 28 October was the frost.

The seven things a page controls

The rewrite above is the most valuable line on the page, and the twenty seconds after it is what they do and do not affect. The outline under both is an outline a reader and a crawler both use.

The seven controls

The same page, with the parts a search engine reads marked

Seven controls connect on one guide, though Search Console still cannot tell which of the extra 270 clicks wanted a date or syrup.

thornfieldbees.co.uk/guides/autumn-feeding
  1. 01When to stop feeding bees in autumn
  2. 02First frost in Shropshire, 2025: 28 October
  3. 03A full super, filed as autumn-feeder-thornfield.jpg
  4. 04moving hives in winter
  5. 05Article markup in the head
  1. 01The h1 and outlineOne h1 on the page, paired with headings that mirror the reader's skim.
  2. 02The answer, above the foldThe base of the pyramid, and the only control a crawler cannot grade.
  3. 03The photographFilename helps a little; descriptive alt text does the real work.
  4. 04An internal linkLeaves this page to make a claim about how pages relate across the site.
  5. 05Structured dataInvisible on the page, visible in the listing, resting on index storage first.

Clicks rose from 210 to 480 over three months, leaving 270 extra visits whose exact intent (date vs syrup) informs the next title.

The base of the pyramid is what separates a page that ranks from one that exists, and it is the only item here that cannot be audited. Pointing at other pages on purpose is sending authority where you want it.

The photograph is alt text, filenames and image search, and the tip of the pyramid is telling search engines what the page is. None of the seven matters at all until the page is stored, which is the mechanism, before any tactic makes sense, and the anchor text above leaves this page and lands in how pages relate, which decides what a site is about.

The 210 and the 480 are one page over three months at one supplier. What Thornfield cannot tell from Search Console is how many of the extra 270 wanted a date and how many wanted syrup, and the answer decides whether the next title should promise a date or a price.

Inside this, 7

Alongside this

Above this