Rank your local SEO fixes by effort, for the week you actually have
A triage prompt for local SEO. It sorts the work into three effort tiers, states a mechanism instead of a percentage, and ranks nothing you did not describe.
- Works in
- ChatGPT, Claude, Gemini
- You need
- What is already done: profile, reviews, pages, listings · The hours you have this month and who can do the work · What you cannot change: budget, developer access, premises · What you have already tried that did nothing
- Written for
- how to improve local seo
Scored by our own engine
This page, run through the audit we sell. Measured 4 August 2026.
Every local SEO article gives you the same list of twenty things. The list is not the problem. The problem is that you have an afternoon, half of the list needs a developer you do not have, and nothing tells you which three items are worth the afternoon. The prompt below sorts the work into three effort tiers, states a mechanism instead of a statistic, and refuses to rank anything whose current state you did not describe.
Triage is not planning
Planning decides what should exist. Triage decides what to do next with the hours you actually have. The two need different inputs, and the input triage needs is the current state.
A model that does not know your primary category is set, or that you have eleven reviews and none this year, will produce the generic list back at you in a nicer order. Rule 1 is what stops that. Anything you did not describe goes into a section called cannot rank, along with the one check that would settle it, which is a more useful answer than a confident guess about a business it cannot see.
Why it will not give you a percentage
Because nobody measured one for your business. This category of advice is full of statistics with no source attached, quoted long enough that they read as facts, and a model asked to justify a ranking will reach for them.
The mechanism sentence replaces the number. It has to say which part of local search an action touches and why, which is a claim you can check. “Improves relevance signals” is padding and you should delete the line. “Your primary category is the field the local results filter on, and yours currently names a service you no longer offer” is a mechanism, and you can act on it today.
Effort is knowable, effect is not
Effort is the half of the trade a model can estimate honestly, because updating a field takes minutes and rebuilding a template takes days regardless of who you are. Effect depends on your competitors, your distance from the searcher and your review history, none of which it can see.
So read the tiers as a cost list that has been sorted sensibly, not as a forecast. Finish everything under an hour before you start anything that takes a day. Then check the cheapest website item on the list, your page titles, with the title tag checker, because templates overwrite that field quietly and it is the one place where an hour of work reliably changes what a searcher reads.
If the triage keeps returning work you have not scoped yet, the problem is upstream: use the local SEO plan prompt to decide what should exist, and the local keyword sets prompt to work out how few pages that turns into.
You are triaging local SEO work for a business with limited time. I already
know roughly what needs doing. What I need is the order, judged on the effort
each item really costs against how much it plausibly moves local visibility.
Rules you must follow:
1. Rank only against the current state I describe. If I have not told you the
state of something, do not assume it is missing and do not assume it is
done. Put it in the section for things you cannot rank.
2. Never state a percentage, a multiple, a number of searches, a traffic
figure or a share of anything. No claims about how many local searches do
one thing or another. Where effect matters, describe the mechanism in one
sentence and let me judge the size.
3. Give effort in real units of work: minutes, hours or days, and the person
who has to do it. "Quick win" is not an effort estimate.
4. Do not recommend anything I told you I already tried, unless you say
precisely what would be different this time.
5. Keep Google Business Profile work, website work and off site work
separate within each tier, because they are done by different people and
block on different things.
6. Do not promise a position, a timeline or a result. You cannot see my
rankings, my competitors or the local results for any query.
7. Where two actions look similar in effect, rank the reversible one higher.
I would rather be able to undo a mistake than optimise one.
Output in three tiers and nothing else. No table.
TIER 1: under an hour each.
TIER 2: about a day each.
TIER 3: weeks, or ongoing.
Inside each tier, order the actions by how much you expect them to matter,
most first. Write every line in exactly this form:
Action. / Effort: the real unit of work. / Mechanism: which part of local
search this touches and why. / Who: me, or the person I do not have. /
Proof: what I would look at to know it did anything.
Then two closing sections:
NOT THIS MONTH. Things commonly recommended for local SEO that I should skip
given what I told you, with the reason for each. Be specific about my
situation rather than listing generic bad advice.
CANNOT RANK THESE. Anything whose priority depends on something you cannot
see or I did not tell you, with the single check that would settle it.
Constraints:
- If the current state I gave you is too vague to order the work, say so and
ask for the three facts that would fix it, instead of ranking against a
guess.
- Do not tell me to "optimise" anything. Name the object and the change.
- If the most useful thing I could do is not an SEO task at all, say that
plainly and put it at the top.
My business and where it operates: [WHAT YOU SELL, TO WHOM, AND WHERE]
What is already done: [PROFILE, REVIEWS, PAGES, LISTINGS, IN WHATEVER DETAIL YOU HAVE]
Time and people: [HOURS THIS MONTH AND WHO CAN DO THE WORK]
Hard constraints: [BUDGET, DEVELOPER ACCESS, PREMISES, ANYTHING FIXED]
Already tried, no result: [WHAT YOU DID AND WHEN, OR "nothing yet"]What to change
Everything in square brackets is yours to replace. Nothing else needs editing.
[WHAT YOU SELL, TO WHOM, AND WHERE]- One sentence with a buyer and a place in it. "We are a two van locksmith covering north Leeds, mostly emergency lockouts for landlords" ranks differently from "locksmith in Leeds", because emergency work makes profile hours and response speed matter more than any page you could write.
[PROFILE, REVIEWS, PAGES, LISTINGS, IN WHATEVER DETAIL YOU HAVE]- The input the whole ranking turns on. Write what is true: "profile claimed, primary category set to Locksmith, 11 reviews, none in the last year, one services page, no hours listed for weekends". Anything you leave out lands in the cannot rank section, which is the honest outcome and better than the model assuming you have nothing.
[HOURS THIS MONTH AND WHO CAN DO THE WORK]- The reason this is triage rather than a plan. "Six hours, me, and a web developer I can email twice" produces a completely different order from "two days a week and an agency", because half the standard local SEO advice needs someone with access to the site template.
[BUDGET, DEVELOPER ACCESS, PREMISES, ANYTHING FIXED]- The things that are not going to change, so nothing gets recommended that you cannot do. "No budget, site is on a builder platform I cannot edit templates on, we work from a home address we do not want published" removes several standard recommendations before they waste a line in the output.
[WHAT YOU DID AND WHEN, OR "nothing yet"]- Stops the output being a list of things you already know did nothing. "Paid for 40 directory listings last spring, no change" is worth writing down, because rule 4 then forces the model either to leave it out or to say what would be different this time. That answer is usually more informative than the ranking.
How to run it
- 01Write down what is already true before you open the model
Spend ten minutes listing the current state: whether the profile is claimed, the primary category, roughly how many reviews and how recent, which pages exist, which directories you are in. This is the only input that makes an ordered list better than a generic one. Anything you cannot state honestly should be left out so it lands in the cannot rank section.
- 02Be honest about the hours
Write the real number of hours you will give this in the next month, not the number you would like to. A ranked list built against two days a week is useless to someone with an afternoon, because the top of it will be work that never starts. Naming the person who can do each thing matters just as much as the time.
- 03Run it and read the last two sections first
Paste the prompt with all five inputs into ChatGPT, Claude or Gemini. Then skip to NOT THIS MONTH and CANNOT RANK THESE before you read the tiers. Those two sections are where a triage prompt earns its keep, because knowing what to leave alone frees more hours than any single fix will.
- 04Check the mechanism on every line you plan to do
Read the mechanism sentence for each action you are about to take and ask whether it describes a real path from the change to a searcher. Vague mechanisms like "improves relevance signals" mean the model is padding. Delete those lines, because an action you cannot explain is an action you cannot judge afterwards.
- 05Do the whole of tier one before touching tier two
Tier one is the set of changes that cost under an hour each, and finishing all of it is nearly always worth more than starting one large project. It also gives you a clean baseline, since making a dozen small changes and one big one in the same week means you learn nothing about either.
- 06Fix the titles on your location and service pages
Title tags are usually the cheapest website side item on the list and the one people leave for last. Run your main service and location pages through the title tag checker to see what actually renders, since templates and plugins rewrite this field quietly. A page whose title is your business name and nothing else is a free improvement sitting in tier one.
Questions people ask
How do I improve local SEO quickly?
Start with the things that cost under an hour and touch what a searcher sees: the primary category on your Google Business Profile, opening hours that are actually correct, a phone number that matches everywhere, and page titles that name the service and the place. These are quick because they are edits rather than projects, and they are worth doing first because they remove the errors that make everything else work less well.
What matters most for local search?
Local results are broadly decided by relevance, distance and prominence. Distance is not something you can edit, which leaves relevance, meaning the words and categories that describe you, and prominence, meaning reviews and how well known you are. That is why a triage list tends to put profile accuracy and reviews above anything you could write on your website.
How is this different from building a local SEO plan?
A plan decides what should exist. This decides what to do on Tuesday. If you have no location strategy yet, start with the local SEO plan prompt, which works out whether a location page deserves to exist at all before anything gets written. Come back here when you already know the work and the problem is that there is more of it than there are hours in the month.
Do directory listings and citations still help?
Consistency helps more than volume. A name, address and phone number that disagree across the places you are already listed is a real problem worth fixing. Buying hundreds of new listings in directories nobody reads is the classic version of effort without effect, which is why the prompt asks what you already tried before it recommends anything.
Can ChatGPT tell me whether my local SEO is working?
No. It cannot see the local pack, your rankings, your competitors or your review counts, and asked anyway it will produce confident numbers that were never measured. Use it to order the work and to say what evidence would prove each change did something. Read the evidence yourself in your profile insights and in Search Console.
Unsubscribe in one click. We never pass your address on.