A blog post now has two readers. A person deciding within four seconds whether to stay, and a machine deciding whether your paragraph is the clearest available answer to a question somebody asked an assistant.
The good news is that they want overlapping things: a direct answer near the top, honest structure, specifics rather than padding. The bad news is that most blog writing advice from the last decade optimised for neither. It optimised for word count.
Quick Info
Start with the question, not the keyword
Write down the exact question a reader would type or ask out loud. Then write down the two questions they will have immediately after you answer it. That trio is your article.
Keyword tools are still useful for volume and for spotting phrasings you would not have guessed, but they will not tell you what the person actually wanted. The People Also Ask box and the follow-up suggestions in AI answers will.
If you cannot state your article as a question, you do not yet know what it is about.
Check what intent the results are serving
Search your question and look at what is already ranking. If the results are all product pages, an article will not rank there no matter how good it is. If they are all listicles, the format is a signal about what readers respond to, not a rule you have to follow.
Read the top three properly. Note what they all cover, which is the baseline you must meet, and what none of them cover, which is your reason to exist.
Outline before you write a sentence
My outline is a list of H2 headings that a reader could read on their own and get the gist of the whole article. If the headings alone do not make sense, the article will not either.
Answer in the first hundred words
The old shape was context, credentials, a story, then the answer at the halfway mark. That shape now loses both audiences. Readers bounce and AI systems extract from whichever paragraph answers the question, which may be a competitor's.
So state the answer plainly in the opening, then earn attention by explaining the nuance, the exceptions and the how. You lose nothing. Anybody who only wanted the answer was going to leave anyway.
Write in extractable units
This is the one structural change I would make to almost every draft I am sent. Write so that any single paragraph could be lifted out and still make sense.
This is what people mean by writing for answer engines. It is not a trick. It is just clear writing with the vague pronouns removed.
Add the thing only you have
Everything above makes an article competent. Competent is now free. What makes it worth citing and worth reading is evidence somebody else cannot produce.
Original evidence is the only part of an article that cannot be summarised out of existence.
Edit for compression, not length
My editing pass has one goal: remove words without losing meaning. Introductions that explain what the article will explain, transitions that restate the previous section, hedges that soften a claim you believe. All out.
A 1,400-word article with nothing removable outranks a 2,600-word article padded to hit a target, because readers stay and because the answer is easier to find.
Finish the technical layer
Add article schema if your platform makes it easy. It helps machines parse authorship and dates. It is not a requirement for being cited and no special file or markup will get you into AI answers on its own.
Publish, then revisit at ninety days
Open Search Console and look at the queries the post is appearing for. Some will be ones you did not target and are easy wins: add a section that answers them properly. Some will show you got the intent wrong, which is a rewrite rather than a tweak.
Posts I update at three and twelve months consistently outperform posts I publish and abandon, by a wide enough margin that I now budget editing time as part of publishing.
Frequently Asked Questions
Long enough to answer the question and its obvious follow-ups, which is usually 1,200 to 2,200 words for a substantial topic. There is no length that ranks; there is coverage that satisfies.
No. There is no file or schema that grants inclusion. Clear structure, a direct answer, and content the system can verify against other sources is what gets you quoted.
No. Use the phrasing your reader would use, in the title, an early paragraph and a heading if it fits naturally. Counting is a habit from a much older version of search.
Only if you can add something a summary cannot: a comparison, a tool, real data, or a strong point of view. Otherwise pick a question further along the decision, where people still click.
Three to five is plenty, chosen because they genuinely help the reader. Link from the specific post to the broad guide and back again; that pairing does most of the work.
For outlines, research and editing, yes, and it saves real time. For the finished prose it strips out the specifics and the voice, which are the only parts doing competitive work.
Before You Go
The process above is not complicated and it is not fast. Four to six hours per post, most of it in outlining and editing rather than drafting. What you get for that is an article that answers a real question better than the alternatives, which is the only durable ranking strategy I have seen survive an algorithm update.
If you want the structural side in more depth, writing content that gets cited by ChatGPT, Gemini and Perplexity goes further, and the headline guide covers the part that decides whether any of this gets read.
Answer first. Prove it. Cut the rest.
I'm a marketing consultant, entrepreneur and content creator. I help businesses grow through practical marketing, websites, SEO, content and AI.
More About Tariq →