What a search engine reads on a page

A search engine reads the text that is in the web page, the headings that structure it, the words used as links, and the alternative text attached to images. That is close to the whole list, and it is very nearly the same list as the one a screen reader can use — which is why accessibility work and search legibility work overlap far more than either is usually sold as. Somewhere on the internet, every one of these web pages is competing with millions of others for the same crawler’s limited attention, which is a separate reason legibility matters even before ranking enters the discussion.

The failures are not exotic. They are ordinary decisions, usually made for visual reasons, that put real content somewhere a machine cannot get at it.

How a search engine actually reaches this content: crawling and indexing

Before a search engine can read any of what follows, it first has to fetch the page at all. A crawler — sometimes called a spider, a term from the same web-as-a-network metaphor — requests each page on a site and follows the links it finds to discover more of them; Google’s crawler is the one that matters most for an Australian small-business site, though Bing and other search engines run their own. What a crawler retrieves then goes through indexing, the process of storing and organising that content so it can actually be returned in a search result. A robots.txt file or a page-level robots directive can block a crawler from fetching or indexing a page entirely — a separate, earlier failure mode from the legibility problems on this page, since a page correctly built for legibility is still invisible if it is blocked from being crawled or indexed in the first place.

What is legible and what is not

Element	Legible	Note
Body text in the page	Yes	The default and the reliable case
Headings	Yes	Structure as well as content
Link text	Yes	Describes the destination
Image alternative text	Yes	Describes the image, not the page
Text inside an image	No, not reliably	Includes text in a design graphic
Text in a video	No	Unless there is a transcript
Text in an embedded PDF	Partly	Indexed separately, outside the site’s structure
Content added by a script	Later and less reliably	Depends on rendering
Content behind a click	Usually yes if it is in the page	Not if it is fetched on demand
Content behind a login	No	By design

Headings

Headings do two jobs at once: they tell a reader what a section is about, and they express the hierarchy of the page. The second is the one that gets damaged.

A heading level chosen because it looked the right size is a heading that misrepresents the structure. A page whose sections are all top-level headings has no hierarchy; a page that skips from a first-level heading to a third has a gap. Both are legible to a machine as a claim about structure, and both make the claim wrongly.

The rule is simple and rarely followed: choose the level by what the section is, and change how it looks in the styling. The Web Content Accessibility Guidelines treat heading structure as an accessibility matter as well, and the Australian obligations that sit behind that are on accessibility obligations.

Text inside images

A price list rendered as a graphic, a service list inside a designed banner, an opening-hours panel that is part of a photograph: all invisible.

This is not usually a decision anyone made. It is what happens when a designer supplies a composed image and it is placed as-is. The cost is that the most commercially specific content on the page — prices, services, hours — is the content most likely to end up inside an image, because it is the content most likely to be designed.

Alternative text is not the fix. Alternative text describes the image for someone who cannot see it, and it is short by nature. Putting a page’s content into it is a misuse that serves neither purpose.

Content a search engine only finds after a script runs

A page whose text is inserted by a script is legible in principle and unreliable in practice: it has to be rendered before the text exists, rendering happens later than fetching, and it can fail for reasons the site owner never sees.

The practical position for a small business site is that commercially important content — the service description, the price, the location, the phone number — should be present in the page as delivered, not assembled afterwards. Whether that is true of a given build is a question for whoever is building it, and it is worth asking, because the answer is usually determined by the platform rather than by a decision.

Link text

The words used as a link describe the destination to both a person and a machine. “Click here” and “read more” describe nothing, appear identically dozens of times on a site, and are listed out of context by screen readers, where they become a list of identical entries.

Descriptive link text costs nothing and is a decision made while writing rather than afterwards. This site enforces it mechanically for that reason.

Tables, lists and other structure

Content that has a shape should carry that shape. A comparison expressed as a table is legible as a comparison; the same content as a paragraph is legible as a paragraph. A list of steps as a numbered list is legible as a sequence.

This matters more than it used to, because extracted answers — the ones that appear above results, and the ones an assistant produces — are drawn from content whose structure makes the relationship explicit.

The overlap with accessibility

Almost everything above is also an accessibility requirement, arising in Australia from the Disability Discrimination Act rather than from any search consideration. Text as text, headings in order, alternative text that describes images, link text that describes destinations, structure expressed as structure.

The overlap is close enough that a site built for accessibility is largely legible by default, and a site with legibility problems usually has accessibility problems in the same places. That is the strongest practical argument for doing the accessibility work, and the accessibility self-assessment is a way to find out where you stand.

What to check on your own site

Select the text on your home page with a mouse and drag. Anything that does not highlight is not text. That single test finds most image-text problems in about ten seconds.

Then read your headings in order, ignoring the content. They should read as an outline of the page. If they do not, the structure is decorative.

What to do next

Add “text is text” to the design brief rather than to the launch checklist, because by launch the images have been made. If you are commissioning writing as a separate piece of work, website copywriting covers where that boundary usually sits.